NEERENGI - Aarhus Universitet

166
NEER ENGI MODELLING AND REDUCING GAS EMISSIONS FROM NATURALLY VENTILATED LIVESTOCK BUILDINGS Biological and Chemical Engineering Technical Report BCE-TR-2

Transcript of NEERENGI - Aarhus Universitet

Page 1: NEERENGI - Aarhus Universitet

NEER ENGI

MODELLING AND REDUCING GAS EMISSIONS FROM NATURALLY VENTILATED LIVESTOCK BUILDINGS Biological and Chemical Engineering Technical Report BCE-TR-2

Page 2: NEERENGI - Aarhus Universitet

DATA SHEET Title: Modelling and reducing gas emissions from naturally ventilated livestock buildings Subtitle: Biological and Chemical Engineering Series title and no.: Technical report BCE-TR-2 Author: Wentao Wu Department of Engineering – Biological and Chemical Engineering, Aarhus University Internet version: The report is available in electronic format (pdf) at the Department of Engineering website http://www.eng.au.dk. Publisher: Aarhus University© URL: http://www.eng.au.dk Year of publication: 2012 Pages: 151 Editing completed: September 2012 Abstract: Livestock buildings are identified to be a major source of ammonia emissions. About 30% of the total ammonia emission within livestock sectors is from naturally ventilated dairy cattle buildings. The main objectives of this study are to predict emissions from naturally ven-tilated dairy cattle buildings and to establish a systematic approach to curtail the emissions. Gas concentrations were measured inside two dairy cattle buildings in mid-Jutland, Denmark. CO2 balance method was thus applied to esti-mate ventilation and emission rates. Computational fluid dynamics (CFD) was used to find the optimum gas sampling positions for outlet CO2 concentration. The gas sampling positions should be located adja-cent to the openings or even in the openings. The NH3 emission rates varied from 32 to77 g HPU-1 d-1 in building 1 and varied from 18 to30 g HPU-1 d-1 in building 2. Scale model experiment showed that partial pit ventilation was able to remove a large portion of polluted gases under the slatted floor. In the full scale simulations, a pit exhaust with a capacity of 37.3 m3 h-1 HPU-1

may reduce ammonia emission only by 3.16% compared with the case without pit ventilation. When the external wind was decreased to 1.4 m s-1 and the sidewall opening area were reduced to half, such a pit venti-lation capacity can reduce ammonia emission by 85.2%. The utilization of pit ventilation system must be integrated with the control of the natu-ral ventilation rates of the building. Keywords: Biotechnology; Computer fluid dynamics modeling; Emission; Environmental engineering; Housing systems; Ventilation and indoor climate Supervisor: Guoqiang Zhang; Peter V. Nielsen; Peter Kai Please cite as: Wentao Wu, 2012. Modelling and reducing gas emis-sions from naturally ventilated livestock buildings. Department of Engineering, Aarhus University. Denmark. 151 pp. - Technical report BCE-TR-2 Cover image: Wentao Wu ISSN: 2245-5817 Reproduction permitted provided the source is explicitly acknowledged

.

Page 3: NEERENGI - Aarhus Universitet

MODELLING AND REDUCING GAS EMISSIONS FROM NATURALLY

VENTILATED LIVESTOCK BUILDINGS

Wentao Wu

Aarhus University, Department of Engineering

Abstract Livestock buildings are identified to be a major source of ammonia emissions. About 30% of the total ammonia emission within livestock sectors is from naturally ventilated dairy cattle buildings. The main objectives of this study are to predict emissions from naturally ventilated dairy cattle buildings and to establish a systematic approach to curtail the emissions. Gas concentrations were measured inside two dairy cattle buildings in mid-Jutland, Denmark. CO2 balance method was thus applied to estimate ventilation and emission rates. Computational fluid dynamics (CFD) was used to find the optimum gas sampling positions for outlet CO2 concentration. The gas sampling positions should be located adjacent to the openings or even in the openings. The NH3 emission rates varied from 32 to77 g HPU-1 d-1 in building 1 and varied from 18 to30 g HPU-1 d-1 in building 2. Scale model experiment showed that partial pit ventilation was able to remove a large portion of polluted gases under the slatted floor. In the full scale simulations, a pit exhaust with a capacity of 37.3 m3 h-1 HPU-1 may reduce ammonia emission only by 3.16% compared with the case without pit ventilation. When the external wind was decreased to 1.4 m s-1 and the sidewall opening area were reduced to half, such a pit ventilation capacity can reduce ammonia emission by 85.2%. The utilization of pit ventilation system must be integrated with the control of the natural ventilation rates of the building.

Page 4: NEERENGI - Aarhus Universitet

i

Preface

This thesis is submitted as a partial fulfillment of the requirements for the Doctor of Philosophy

(PhD) degree at Department of Engineering, Faculty of Science and Technology, Aarhus

University, Denmark. The research was performed between June 2009 and June 2012 supervised by

senior scientist Guoqiang Zhang. Scale model measurements were conducted in the wind tunnel at

Air Physics Lab, Research Centre Bygholm, Horsens. Laboratory experiments using small wind

tunnels were carried out at Air Physics Lab, Research Centre Foulum, Tjele. Full scale field

measurements on gas emissions were completed in two naturally ventilated dairy cow buildings in

Jutland. Part of the numerical simulations was done in the University of Colorado at Boulder, USA.

This thesis is based on the work presented in three published research papers and three

manuscripts. They are:

1. Wu, W., Zhang, G., Kai, P., 2012. Ammonia and Methane Emissions from Two Naturally

Ventilated Dairy Cattle Buildings and the Influence of the Climatic Factors on Ammonia

Emissions. Atmospheric Environment 61, 232-243.

2. Wu, W., Zhang, G., 2012. Turbulent Air Characteristics in Two Naturally Ventilated Dairy

Cattle Buildings. Submitted to a peer review journal.

3. Wu, W., Zhai, J., Zhang, G., Nielsen, P.V., 2012. Evaluation of methods for determining air

exchange rate in a naturally ventilated dairy cattle building with large openings using

computational fluid dynamics (CFD). Submitted to a peer review journal.

4. Wu, W., Kai, P., Zhang, 2012. An assessment of a partial pit ventilation system to reduce

emission under slatted floor-Part 1: Scale Model Study. Computers and Electronics in

Agriculture 83, 127-133.

5. Wu, W., Zhang, G., Bjerg, B., Nielsen, P.V., 2012. An assessment of a partial pit ventilation

system to reduce emission under slatted floor-Part 2: Feasibility of CFD prediction using

RANS turbulence models. Computers and Electronics in Agriculture 83, 134-142.

6. Wu, W., Zong, C., Zhang, G., 2012. Large eddy simulation of airflow and pollutant transport

under slatted floor of a slurry pit model. Submitted to a peer review journal.

During this study, by collaborating with other scientists, the following publication was

produced:

1. Saha, C.K., Wu, W., Zhang, G., Bjerg, B., 2011. Assessing effect of wind tunnel sizes on air

velocity and concentration boundary layers and on ammonia emission estimation using

computational fluid dynamics (CFD). Computers and Electronics in Agriculture 78 (1), 49-

60.

Page 5: NEERENGI - Aarhus Universitet

ii

Acknowledgements I wish to acknowledge the following persons who have contributed to my pursuing of PhD in

diverse manners.

My first and foremost appreciation goes to my principal supervisor, Dr. Guoqiang Zhang,

Senior scientist in Department of Engineering, Aarhus University. His broad knowledge and

enthusiastic attitude on science always inspired me on my work. His great personality led me to

shape myself in the very same fashion.

I would like to extend my appreciation to my co-supervisor - Professor Peter V. Nielsen from

Aalborg University, project supervisor - Dr. Peter Kai from AgroTech, Denmark. Many thanks are

given to Associate professor Bjarne Bjerg from Copenhagen University, Associate professor John

Zhai from University of Colorado at Boulder, USA. When I stayed in these two universities, they

advised me on Computational Fluid Dynamics (CFD) and my skills on CFD were quantitatively

improved since then. I would like to express my gratitude to Dr. Chayan Kumer Saha from

Leibniz-Institut für Agrartechnik Potsdam-Bornim e.V., Germany and PhD student Raphael

Kubeba from Harper Adams University College, UK. They read through my thesis and helped me

to improve it. I benefited from the cooperation with PhD student Chao Zong at Aarhus University.

I deliver my gratitude to the laboratory technicians Jan Ove Johnsen and Jens Kristian

Kristensen at Aarhus University. The PhD project would not be possibly completed without their

large contributions. I would like to thank Anja Torup Hansen, who helped me since I was on the

way to Denmark. I am always feeling grateful to Dr. Li Rong at Aarhus University, Professor Jing

Liu, Senior Engineer Xiaoxin Man from Harbin Institute of Technology, China. Their help or

support enabled me to come to Denmark.

My special appreciation goes to the greatest and most beautiful country in this world –

Denmark. In this ever loving land, I met several good friends – Dr. Jana Boleckova from Czech

Republic, Sebastiano Falconi from Sardinia, Emmanuel Arthur from Ghana, Miroslaw Marek

Kasprzak from Poland, Grazina Kadziene from Lithuania and Kristin Barkve Andersen from

Norway. The friendship with Jana and Sebastiano was one of the most precious gifts for the three

years in Denmark. Pernille Bildsøe and Ole Jørgensen never hesitated to translate Danish to

English for me in diverse fashions. I can never stop writing this list. In brief, I gave my sincere

thanks to all the people being helpful and kind to me.

Finally, I express my gratitude and love to my grandma and my wife. My grandma brought

me up and provided me the opportunity to experience the wonderful trip in the life. The delicious

dinner provided by my wife Lei Wang always erased the tiredness of the whole work day. This

work is also dedicated to my grandma and my wife.

(Wentao Wu)

Page 6: NEERENGI - Aarhus Universitet

iii

Summary

Emissions of ammonia to the atmosphere cause acidification of soil and increase eutrophication to

sensitive ecosystems such as aquatic systems, while emissions of greenhouse gases such as nitrous

oxide, carbon dioxide and methane influence global climate. Livestock buildings are identified to be

a major source of those gas emissions. About 30% of the total ammonia emission within livestock

sectors is from naturally ventilated dairy cattle buildings. Nearly no innovative low emission

technology has been applied to cattle buildings. Besides mitigation techniques, methods to quantify

gas emissions from naturally ventilated livestock buildings are still lacking. Therefore, the main

objectives of this study are to predict emissions from naturally ventilated dairy cattle buildings and

to establish a systematic approach to curtail the emissions.

Gas concentrations were measured inside two dairy cattle buildings in mid-Jutland, Denmark.

CO2 balance method was thus applied to estimate ventilation and emission rates. To evaluate the

influence of climatic parameters on gas emissions, air velocity, turbulence and temperature were

monitored inside and outside the buildings. Computational fluid dynamics (CFD) was used to find

the optimum gas sampling positions for outlet CO2 concentration. A concept of partial pit

ventilation system was proposed to abate gas emissions. The hypothesis was verified first by a 1:2

scale model in wind tunnel measurements. Since a partial pit ventilation system cannot be found in

practical cattle buildings, the investigation on the potential of such a system to reduce emission was

carried out by CFD simulations.

The NH3 emission rates varied from 32 to77 g HPU-1 d-1 in building 1 and varied from 18

to30 g HPU-1 d-1 in building 2. The average emission of CH4 was 290 and 230 g HPU-1 d-1 from

building 1 and 2, respectively. Diurnal pattern was found for NH3 and CH4 emission rates. From

multiple linear regression models, there was a significant linear relationship between NH3 emission

rates and climatic factors including the external wind speed as well as the air temperature

(P<0.001), but not with the external wind directions (P>0.05).

Air exchange rates (AER) predicted by integration of volume flow rate (VFR) and tracer gas

decay (TGD) were in good agreement with each other within a large range of wind speeds. Large

difference in AER estimation was found between VFR and constant tracer gas method (CTG) using

the mean CO2 concentration of the entire room as outlet concentration. It indicates that the mean

CO2 concentration of the entire room may not represent the outlet gas concentration. The gas

sampling positions should be located adjacent to the openings or even in the openings. All the

openings especially of different azimuths should possess sampling tubes. The maximum gas

concentrations at different openings could be the optimum value to represent the concentration in

the exit air.

Scale model experiment showed that partial pit ventilation was able to remove a large portion

of polluted gases under the slatted floor. RANS turbulence models especially RSM can be used to

Page 7: NEERENGI - Aarhus Universitet

iv

predict the removal capability of a partial pit ventilation system to reduce emission under slatted

floor. In the full scale simulations, a pit exhaust with a capacity of 37.3 m3 h-1 HPU-1 may reduce

ammonia emission only by 3.16% compared with the case without pit ventilation when the external

wind was 4.2 m s-1. When the external wind was decreased to 1.4 m s-1 and the sidewall opening

area were reduced to half, such a pit ventilation capacity can reduce ammonia emission by 85.2%.

The utilization of pit ventilation system must be integrated with the control of the natural ventilation

rates of the building.

The study is helpful for emission inventory and also demonstrates that partial pit ventilation

system may be a feasible approach to mitigate emission from naturally ventilated cattle buildings.

Page 8: NEERENGI - Aarhus Universitet

v

Resumé på Dansk

Ammoniak emission til atmosfæren, forsager forsuring af jord og forårsager eutrofiering af

kvælstof-sensitive økosystemer såsom vandløb. Udledning af drivhusgasser såsom N2O, kuldioxid

og metan påvirker klimaet globalt. Landbrugets dyrehold bidrager væsentligt til drivhusgasser i det

global klima regnskab. Omkring 30% af de samlede ammoniakemissioner fra dansk landbrug,

stammer fra ventilation af kvægbygninger. Hidtil har anvendelsen af innovative teknologier til

reduktion af emission været begrænset, dertil mangler der metoder til kvantificering af

emissionsniveauer. Formålet med dette studie har derfor været at fastlægge emissioner fra

bygninger med dyrehold, samt at finde en systematisk metode hvormed emissionerne kan

begrænses.

I bestræbelserne på at fastlægge emissioner fra stalde med malkekvæg blev

ventilationsluftmængden fastlagt indirekte ved hjælp af CO2 balancemetoden.

Computersimuleringer(CFD) blev brugt til at finde optimale placeringer af gas-sensorer for at

anvende CO2 massebalance metoden.

Et system med lokal luftudsugning i gyllekanalen, blev forslået med henblik på at begrænse

gasemissionen. Tesen blev først afprøvet med en 1:2 skala model i vindtunnel forsøg. Da der ikke

fandtes et eksisterende system til lokal ventilation af gyllekanalen, blev potentialet for emissions

reduktion undersøgt med computer simuleringer.

Emissionen af NH3 varierede fra 32-77 g HPU-1 d-1 i bygning 1 og 18-30 g HPU-1 d-1 i bygning 2.

Det gennemsnitlige emissionsniveau for CH4 for henholdsvis bygning 1 og 2 var 290 og 230 g

HPU-1 d-1. Emissionerne af NH3 og CH4 viste en tydelig døgnvariation. Fra et sæt af flere lineære

regressioner, fandtes en signifikant linear sammenhæng mellem ammoniakemissionen og de

klimatiske parametre, ekstern vindhastighed samt lufttemperaturen(P<0.001), men ikke

vindretningen (P>0.05).

Luftudskiftningshastigheden estimeret ved integration af den volumetriske flow rate (eng.

AER), samt målinger på henfald af sporgasser (eng. TGD), viste god ovenstemmelse over et bredt

interval af vindhastigheder. Store forskelle i luft-udskiftningshastigheden baseret på henholdsvis

sporgas og integrationsmetoden, indikerer at den gennemsnitlige CO2 koncentration i rummet, ikke

svarer til koncentrationen målt i udluftnings- åbningen. Målepostionerne for gassensorer skal

placeres tæt på udluftnings åbninger, eller helst i selve åbningen. Der skal monteres gas-samplinger

i alle åbninger, især i vinklede rør. Den maximale gaskoncentration målt i de forskellige åbninger,

fremgik som værende den optimale værdi til at repræsentere koncentrationen i ventilations luften.

Eksperimenter med skalamodeller af staldbygninger viste, at metoden med lokal udluftning af

gyllekanaler er i stand til at fjerne en stor andel af de forurenende gasser, der produceres og frigives

under spaltegulvet. RANS turnolens modeller og især RSM kan estimere potientialet for reduktion

af gasniveauet ved lokal ventilation under spaltegulvet.

CFD simulationer af en fuldskala kvægstald viste, at luftudsugning under spaltegulvet med en

Page 9: NEERENGI - Aarhus Universitet

vi

kapacitet på 37,3 m3h-1HPU-1 kan give en reduktion i ammoniakemissionen på 3,16% sammenlignet

med en traditionel gyllekanal ved ekstern vindhastighed på 4,2ms-1. Hvis den eksterne

vindhastighed reduceres til 1,4ms-1, og ventilationsåbningsarealet i staldens facader reduceres til det

halve, kan den luftudsugning reducere ammoniakemissionen med 85,2%. For at udnytte et

gyllekanal luftudsugningssystem, er det nødvendigt at designet integreres med bygningens øvrige

ventilation.

Studiet har demonstreret at lokal udluftning under spaltegulvet kan være en mulig løsning til

begrænsning af emisionen af ammoniak fra naturligt ventilerede staldbygninger.

Page 10: NEERENGI - Aarhus Universitet

vii

Table of Contents

Preface ................................................................................................................................................................. i

Acknowledgement ............................................................................................................................................. ii

Summary .......................................................................................................................................................... iii

Résumé på Dansk .............................................................................................................................................. v

Table of contents ............................................................................................................................................. vii

Chapter 1 ........................................................................................................................................................... 1

1.1. Ammonia emissions from naturally ventilated dairy cattle building ................................................. 2

1.2. Available methods to determine natural ventilation rates ................................................................. 3

1.3. The challenge to determine ventilation rates and gas emission rates ................................................ 5

1.4. Modelling naturally ventilated cattle building ................................................................................... 6

1.5. Environmental technologies to curtail emissions .............................................................................. 7

1.6. Research objectives ........................................................................................................................... 8

1.7. Outline of this thesis .......................................................................................................................... 8

References ......................................................................................................................................................... 9

Chapter 2 ..........................................................................................................................................................13

Abstract ............................................................................................................................................................14

2.1. Introduction ......................................................................................................................................15

2.2. Materials and Methods .....................................................................................................................16

2.2.1. Buildings ...................................................................................................................................16

2.2.2. Production and feed ..................................................................................................................19

2.2.3. Measurement setup ...................................................................................................................19

2.2.4. Calculations of ventilation and emission rate ...........................................................................20

2.2.5. Statistical analysis.....................................................................................................................21

2.3. Results ..............................................................................................................................................22

2.3.1. Gas concentrations ....................................................................................................................22

2.3.2. Daily average emission rates of NH3 and CH4 and diurnal variations .....................................24

2.3.3. The effect of the climatic factors on the air exchange rate .......................................................25

2.3.4. The effect of climatic factors on NH3 emission rates ...............................................................27

2.4. Discussion .........................................................................................................................................28

2.4.1. Gas concentrations ....................................................................................................................28

2.4.2. NH3 and CH4 emission rates .....................................................................................................29

2.4.3. The effect of the climatic factors ..............................................................................................30

2.5. Conclusions ......................................................................................................................................31

Page 11: NEERENGI - Aarhus Universitet

viii

References ........................................................................................................................................................32

Chapter 3 ..........................................................................................................................................................35

Abstract ............................................................................................................................................................36

3.1. Introduction ......................................................................................................................................37

3.2. Methods and materials ......................................................................................................................38

3.2.1. Experimental details .................................................................................................................38

3.2.2. Data analysis .............................................................................................................................38

3.2.3. Background of statistical methods ............................................................................................38

3.2.4. Theory of spectral analysis .......................................................................................................41

3.3. Results and discussion ......................................................................................................................42

3.3.1. Hourly averaged air velocities outside and inside the buildings ..............................................42

3.3.2. Sampling period for statistical and spectral analysis ................................................................42

3.3.3. Autocorrelation and integral turbulence length scales ..............................................................43

3.3.4. Turbulence kinetic energy, dissipation rate and small scales ...................................................45

3.3.5. Cross-correlation and power spectrum .....................................................................................48

3.4. Conclusions ......................................................................................................................................53

References ........................................................................................................................................................53

Chapter 4 ..........................................................................................................................................................55

Abstract ............................................................................................................................................................56

4.1. Introduction ......................................................................................................................................57

4.2. Materials and methods ......................................................................................................................58

4.2.1. The field measurement for model validation ............................................................................58

4.2.2. CFD Model details ...................................................................................................................60

4.3. Results and discussion ......................................................................................................................65

4.3.1. Model validation .......................................................................................................................65

4.3.2. Assessing the capability of VFR, TGD and CTG to calculate AER ........................................68

4.3.3. The representative CO2 concentration at the exit air ................................................................69

4.3.4. Further discussion .....................................................................................................................72

4.4. Conclusion ........................................................................................................................................73

References ........................................................................................................................................................74

Chapter 5 ..........................................................................................................................................................77

Abstract ............................................................................................................................................................78

5.1. Introduction ......................................................................................................................................80

5.2. Materials and Methods .....................................................................................................................80

Page 12: NEERENGI - Aarhus Universitet

ix

5.2.1. Experimental facility ................................................................................................................80

5.2.2. Experimental set-up and measurements ...................................................................................83

5.2.3. Estimation of air exchange rate using removal ratio ................................................................84

5.2.4. The integrate variable to evaluate the effect of all factors on the performance of pit ventilation 85

5.2.5. CO2 concentration difference ...................................................................................................85

5.3. Results ..............................................................................................................................................85

5.3.1. CO2 concentration inside the pit headspace ..............................................................................85

5.3.2. The effect of the air velocity .....................................................................................................86

5.3.3. The effect of the slatted floor ...................................................................................................86

5.3.4. The effect of the pit ventilation configurations (Ventilation rate and position) .......................87

5.3.5. The overall effect of the four factors ........................................................................................87

5.4. Discussion .........................................................................................................................................87

5.4.1. CO2 concentration inside the pit headspace ..............................................................................87

5.4.2. The effect of air velocity ..........................................................................................................88

5.4.3. The effect of the slatted floor ...................................................................................................89

5.4.4. The effect of pit exhaust configurations ...................................................................................89

5.4.5. The overall effect of all the factors ...........................................................................................90

5.5. Conclusions ......................................................................................................................................90

References ........................................................................................................................................................91

Chapter 6 ..........................................................................................................................................................92

Abstract ............................................................................................................................................................93

6.1. Introduction ......................................................................................................................................95

6.2. Materials and Methods .....................................................................................................................96

6.2.1. Basic concept of CFD and software .........................................................................................97

6.2.2. Computational domain .............................................................................................................97

6.2.3. Boundary conditions .................................................................................................................98

6.2.4. Mesh .........................................................................................................................................99

6.2.5. Calculation of vertical mean and turbulence flux ...................................................................100

6.3. Results ............................................................................................................................................100

6.3.1. The effect of simplification of computational domains ..........................................................100

6.3.2. The effect of different RANS turbulence models ...................................................................100

6.3.3. Distribution of velocities and concentrations .........................................................................101

6.3.4. Airflow patterns ......................................................................................................................102

Page 13: NEERENGI - Aarhus Universitet

x

6.3.5. Removal ratios predicted by CFD ..........................................................................................103

6.3.6. The vertical CO2 flux through the slatted floor openings .......................................................104

6.4. Discussion .......................................................................................................................................104

6.4.1. The simplification of computational domains ........................................................................104

6.4.2. Velocity fields and effects of different RANS turbulence models .........................................105

6.4.3. Removal ratios predicted by CFD ..........................................................................................106

6.5. Conclusion ......................................................................................................................................107

References ......................................................................................................................................................107

Chapter 7 ........................................................................................................................................................110

Abstract ..........................................................................................................................................................111

7.1. Introduction ....................................................................................................................................112

7.2. Materials and methods ....................................................................................................................113

7.2.1. Measurements for model validation .......................................................................................113

7.2.2. CFD modelling .......................................................................................................................115

7.2.3. Strouhal number .....................................................................................................................118

7.2.4. Retention time ........................................................................................................................119

7.2.5. Vertical mass flux induced by turbulent diffusion and mean flow .........................................119

7.3. Results and discussion ....................................................................................................................119

7.3.1. Model validation .....................................................................................................................119

7.3.2. Flow field and NH3 distribution in the headspace ..................................................................123

7.3.3. Velocity and NH3 fluctuation in the headspace ......................................................................123

7.3.4. Retention time ........................................................................................................................126

7.3.5. The pollutant removal mechanism for the two simulation methods .......................................126

7.3.6. Further discussion and future work ........................................................................................126

7.4. Conclusions ....................................................................................................................................127

References ......................................................................................................................................................128

Chapter 8 ........................................................................................................................................................131

Abstract ..........................................................................................................................................................132

8.1. Introduction ....................................................................................................................................133

8.2. Material and methods .....................................................................................................................133

8.2.1. Geometry and mesh ................................................................................................................133

8.2.2. Boundary conditions ...............................................................................................................134

8.2.3. Turbulence models and numerical methods ...........................................................................136

8.2.4. Validation of the CFD model .................................................................................................136

Page 14: NEERENGI - Aarhus Universitet

xi

8.3. Results ............................................................................................................................................136

8.3.1. Model validation .....................................................................................................................136

8.3.2. Potential of pit ventilation system to reduce ammonia ...........................................................137

8.4. Discussion .......................................................................................................................................139

8.4.1. Model validation .....................................................................................................................139

8.4.2. Potential of pit ventilation system to reduce ammonia ...........................................................139

8.5. Conclusion ......................................................................................................................................140

References ......................................................................................................................................................140

Chapter 9 ........................................................................................................................................................142

9.1. Introduction ....................................................................................................................................143

9.2. Determination of ventilation and gas emission rates ......................................................................144

9.3. Climatic factors on gas emissions ..................................................................................................145

9.3.1. Air velocity and temperature ..................................................................................................145

9.3.2. Air turbulence characteristics .................................................................................................146

9.4. Potential of partial pit ventilation systems to reduce emission .......................................................146

9.4.1. Laboratory measurements .......................................................................................................146

9.4.2. Using CFD to investigate the performance of pit ventilation system .....................................147

9.4.3. Modeling of slatted floor ........................................................................................................147

9.4.4. Full scale simulation ...............................................................................................................148

9.5. Perspectives ....................................................................................................................................148

9.6. General conclusions ........................................................................................................................149

References ......................................................................................................................................................150

Page 15: NEERENGI - Aarhus Universitet

1

Chapter 1

General Introduction

Page 16: NEERENGI - Aarhus Universitet

2

1.1. Ammonia emissions from naturally ventilated dairy cattle buildings

The recognition of the ammonia emissions as an environmental issue can be marked by the work of

Van Breemen et al. (1982), which discovered the fact of soil acidification from atmospheric

ammonium sulphate and identified the sources of the ammonia emissions – the livestock production

units. Inventories have shown that animal housing, stored animal manure and exercise areas account

for about 60-80% of the total emission of NH3 in Europe (Hutchings et al., 2001). The estimated

NH3 emissions from various animal husbandry operations in some countries are shown in Table 1.1.

Ammonia emissions from animal husbandry in Denmark during 1990 – 2005 are shown in Table

1.2.

Table 1.1 Ammonia emission estimates from different animal husbandry in three countries

Source Ammonia emissions(Kg animal-1 year-1) Denmark

(Peterson, 2006) Czech

(Zapletal and Chroust, 2006) USA

(Battye et al., 2003) Dairy cows 13.55 27.9 28 Other cattle 7.99 64.8 10.2 Sows 6.39 17.44 16.4 Pigs 6.24 14.8 6.4 Broilers 0.08 0.21 0.28 Hens 0.40 0.92 0.31

Table 1.2 Ammonia emission estimates from animal husbandry in Denmark during 1990 – 2005

(tones NH3-N) (Gyldenkærne and Mikkelsen, 2007)

Animal 1990 2000 2005 type Emission

s % of the total Emission

s % of the total Emission

s % of the total

emissions emissions emissions Cattle 33,598 41.8 22,201 36.6 16,140 30.0 Horses 1,246 1.55 1,211 1.99 1,174 2.18 Sheep 280 0.34 201 0.34 191 0.32 Pigs 36,139 44.9 28,179 46.4 27,452 51.1 Poultry 4,204 5.23 4,843 7.98 4,581 8.52 Fur animals

4,975 6.18 4,064 6.69 4,237 7.88

Total 80,441 100 60,699 100 53,774 100

The two tables show that the cattle buildings, which are usually naturally ventilated, constitute one

of the largest NH3 emission sources within agricultural production. More than 30% of the total

ammonia emission originated from cattle buildings in Denmark before 2005 and this added weight

to the concern of environmental pollution issues caused by naturally ventilated cattle buildings. The

reliable measurement methods and protocols of ammonia concentrations and air volumes from

Page 17: NEERENGI - Aarhus Universitet

3

naturally ventilated houses were identified as key problems in European commission meeting in

1990. The first problem was the greatest uncertainty in the estimates of the natural ventilation rates.

The second problem was the lack of the reliable measurement technologies to determine the outlet

gas concentration to quantify the emission rates. To date, there has been little innovation for

overcoming the two challenges. Meanwhile, limited application of emission abatement technologies

can be found for ammonia reduction from cattle buildings. In order to find new low emission

technologies to reduce emissions from naturally ventilated cattle buildings, quantification of

ventilation rates and ammonia emission rates is the first key step.

1.2. Available methods to determine natural ventilation rates

For a mechanically ventilated livestock building, ventilation rate can be monitored by a pressure

nozzle equipped to exhaust duct or by recording the speed of the exhaust fan. The method cannot be

applied directly to naturally ventilated buildings which possess large openings. The openings may

act as inlet at one condition and as outlet at another condition according to the external wind

directions. The variability of the opening function requires special methods to determine natural

ventilation rates. In 1978, Strøm developed equations for total and sensible heat productions.

Afterwards, CIGR (1984 & 2002) working group derived further common balance equations for

heat and moisture based on heat production. The three balance equations are:

Heat balance : TcVTAUS rb (1.1)

Moisture balance : HVL r 680 (1.2)

CO2 balance : CVQ r (1.3)

where, Sb is sensible heat production, W; A is the surface area of the livestock building, m2; U is the

heat transmission coefficient for building surfaces, W m-1 K-1, ΔT is the temperature difference

between indoors and outdoors, K; Vr is the ventilation rate, m3 h-1; c is the specific heat of air, J m-3

K-1; L is latent heat production, W; ΔH is the difference in water content between indoor and

outdoor air, kg m-3; Q is the CO2 production by animals, m3 h-1; ΔC is the difference in CO2 content

between indoor and outdoor air, ppm.

The establishment of the three balance equations enables natural ventilation rate to be

calculated via heat or mass balance method. Pederson et al. (1998) compared three methods for

calculation of the ventilation rate in Northern European livestock buildings on the basis of the

balances of heat, moisture and CO2 for dairy cattle; the result showed that a CO2 production of

0.185 m3 h-1 HPU-1 provided good agreement between ventilation rates calculated from CO2

production and those from temperature and moisture; for naturally ventilated cattle buildings, CO2

production method is superior to the other two methods due to the small difference of indoor and

outdoor temperature and moisture in the air. The above methods are only validated for steady state

conditions. Diurnal variation in the production of metabolic heat, water and CO2 will limit the

accuracy of those methods. The other uncertainties of this CO2 balance method can be: (1) a

Page 18: NEERENGI - Aarhus Universitet

4

uniform distribution of CO2 production is assumed, which is unlikely in practical buildings; (2) the

ventilation based on equation (1.1 to 1.3) depends on the measurement location of the indoor

temperature, water vapour and CO2 content. Attempting to overcome uncertainty in estimation of

ventilation rate brought by the heterogeneity of CO2 distribution, Teye and Hautala (2007)

developed theories for open stall dairy building without ideal mixing under strong wind. The mass

balance within a short distance dx along the air flow at position x inside the building is:

rr VdxxCALQdxVxC )(/)( ' (1.4)

where, C denotes the gas concentration, ppm; L is the total length along the air flow, m; A’ is the

total area for CO2 production, m2; Integrating equation (1.4), they obtain: 'CVQ r (1.5)

where, ΔC’ is the gas concentration change within the building instead of the difference between

indoor and outdoor gas concentrations. The equation (1.5) may guide to setup the measurements of

gas concentrations. In other words, the equation can provide theoretical background for grid

measurements with high space resolution to achieve accurate estimation of natural ventilation rates

and gas emissions. However, equation (1.4) is questionable in several aspects. Firstly, the flow and

gas distribution inside a cattle building are three dimensional. Hence, a one dimensional mass

balance along dx cannot be correct. Instead, mass balance should be established based on a three-

dimensional finite volume. Secondly, since the equation intends to consider ideal mixing, the

heterogeneity of CO2 production should also been considered. However, the distribution of CO2

production expressed as Q/A’ is still uniform. Additionally, the ventilation rate for mass balance at a

micro position dx should be local ventilation rate, which changes according to positions. The

assumption of constant local ventilation rate expressed as Vr in equation (1.4) is not acceptable.

Therefore, grid measurement still cannot surely represent the position for sampling indoor

temperature, water vapour and CO2.

The deficiency of the CO2 production method provokes scientist to develop other methods to

calculate ventilation rates. Demmers et al. (2001) used the pressure difference across the ventilation

openings to estimate the ventilation rate. A simple explanation of the method can be illustrated by

the following equation.

2

22

1r

do

VcA

p (1.6)

where, p is the pressure difference, Pa; Ao is the ventilation opening area, m2; Cd is the discharge

coefficient. But the method failed to balance the mass flow rates in and out through all openings of

the building. This could be due to the inadequate information on obtaining Cd and the uncertain of

measuring p . At low wind speeds (<2m s-1), the pressure drop across openings was small and

pressure difference cannot be measured reliably. The method is not practical until a more precise

Page 19: NEERENGI - Aarhus Universitet

5

understanding of the Cd for full scale building and more precise pressure sensor for low pressure

difference measurement were available.

Tracer gas methods including tracer decay and constant tracer injection are widely used to

measure natural ventilation rate. The constant tracer injection can be perceived as an equivalent

method with mass balance method discussed in the above section. The method has the same

deficiency with the mass balance method except that it can generate uniform release of the tracer.

The principle of the tracer gas decay is to dose a quantity of tracer gas in the room and then to

record the decay rate of the tracer concentration. Tracer gas decay was concluded as an appropriate

technique to quantify ventilation rate from naturally ventilated buildings (Snell et al., 2003; Samer

et al., 2011). However, the major shortcoming of the method is that it cannot discover the variation

of the ventilation rate during this decay period. Meanwhile, it is difficult to use the meathod record

ventilation rates continuously with respect to time.

Based on the discussion of the above methods, assessment of the different methods is

necessary in order to find an optimum approach to quantify the natural ventilation rates from dairy

cattle buildings.

1.3. The challenge to determine ventilation rates and gas emission rates

In order to quantify ammonia emission rates using CO2 production model, two factors are prior: the

ventilation rates and the average ammonia concentration in the air leaving the cattle building. The

first barrier for estimating emission rates from naturally ventilated buildings is lacking of an

efficient solution to estimate ventilate rate, which was discussed in section 1.2. The second barrier

is lacking of sufficient knowledge on where to measure the gas concentrations to represent the

outlet concentration, which was termed as representative gas concentration in this study. Much

research has been attempted to develop approaches to find proper sampling positions to measure the

representative gas concentration. Demmers et al. (1998) measured the representative gas

concentration with two sampling systems: nine positions in openings around the perimeter of the

building and a ring sampling line in the building; the ring line gave consistently higher ventilation

rates than the perimeter discrete positions. However, the analysis of the influence of the two

systems was not extended to emission rate while only perimeter samples were used for the

calculation of the emission rate. Zhang et al. (2005), Ngwabie et al. (2009) and Samer et al. (2012)

used the gas concentrations at multiple locations inside the building as the representative

concentration; the uncertainty introduced by sampling locations was not discussed. Feidler and

Müller (2011) recorded gas concentrations at 12 internal points and considered the variability of

emissions by using gas concentration at different positions; they argued that more internal sampling

positions were required to obtain a representative gas concentration. The conclusion seemed

controversial to that drawn by Demmers et al. (1998), who addressed that the internal measuring

positions could lead to high uncertainties.

Page 20: NEERENGI - Aarhus Universitet

6

1.4. Modelling naturally ventilated cattle building

Both the ventilation rate and the distribution of gas concentration depend on the external wind and

internal airflow patterns including air velocities and turbulences. Due to the importance of the

airflow features, devices are required to depict the flow characteristics in the building.

Computational Fluid Dynamics (CFD) can offer effective solutions for flows in both spatial and

temporal fields. Nielsen (1974) first applied CFD for room airflow prediction. The first CFD study

on livestock buildings crudely approximated the representative geometry as a two dimensional

rectangle (Choi et al., 1988). Afterwards, more scientists worked to develop CFD models to

consider complicated geometrical details of livestock buildings (Sun et al., 2004; Bjerg et al.,

2008a; Bjerg et al., 2010). Comprehensive simulations of naturally ventilated cattle buildings were

performed by Norton et al. (2010a) and Norton et al. (2010b). All these work proved that CFD had

the potential to model the naturally ventilated building and to provide concrete flow information for

estimation of ventilation rate.

Issues concerning geometry simplification are always confronted to model naturally

ventilated dairy cattle buildings. The detailed geometrical description including animals and

partitions is difficult to model directly in a CFD simulation. The complexity of the geometry will

require massive meshes and long time to iterate the calculations. The animal occupied zones (AOZ)

and was tackled as porous media in the work of Sun et al. (2004), Bjerg et al. (2008a) and Norton et

al. (2012a, b). The resistance coefficients of porous media need to be set up in the simulation but

there is not enough data to derive the coefficients by measurement. Bjerg et al (2008b) assumed a

uniform distribution of animals and extracted part of AOZ to calculate the pressure drop using

CFD; the coefficients were obtained by a linear regression from the pressure drop and the

associated inlet velocity.

For cattle buildings with slatted floor, two major airflow circulations exist: one is in the room

space and one is in the pit headspace under the slatted floor. So far, some studies regarding airflow

patterns and pollutant dispersion in dairy cow buildings have been limited to the space above the

slatted floor (Norton et al., 2010a; Norton et al., 2010b). However, little attention has been given to

the airflow characteristics under the slatted floor, which is important to understand ammonia

emission from the slurry surface. Therefore, modelling of slatted floor becomes the main concern in

order to illustrate the flow patterns in the pit headspace. The slot width in a real cattle building is

about 0.02 m, while the shortest building dimension is generally longer than 5 m. The small ratio of

slot width and building dimension prevents a direct modelling of the geometrical details. Therefore,

slatted floor is usually tackled as porous media (Sun et al., 2004; Bjerg et al., 2008a; Bjerg et al.,

2008b). However, up to date, the difference between a simulation of slatted floor with geometrical

details and that using porous media cannot be found in literature. The uncertainty of using porous

media should be documented before used in simulation.

Page 21: NEERENGI - Aarhus Universitet

7

1.5. Environmental technologies to curtail emissions

The aim of integration of knowledge on estimating ventilation rate and emission rate is to find

solutions to curtail gas emissions from naturally ventilated dairy cow buildings. Measures to reduce

the ammonia emission can be achieved by adjusting or regulating factors influencing ammonia

emissions. These factors include:

Diet including intake of crude protein content and feeding schedule (Philippe et al., 2011).

Livestock type. If the animal is cow, milk yield also affects ammonia emission.

Housing system such as slatted floor type, ventilation control, etc (Philippe et al., 2011).

Manure removal system (Philippe et al., 2011).

Meteorological and indoor climatic conditions: ambient temperature and relative humidity

(Philippe et al., 2011), air velocity and turbulence intensities (Saha et al., 2010b).

Altering the diet is regarded as an effective and direct way of achieving reduction of nitrogen

excretion in urine. Smits et al. (1995), Elzing and Monteny (1997) found approximately 40%

ammonia emission can be abated by feeding cows with low-N diet instead of high-N diet.

Possibilities for the reduction of the nitrogen content of the diet were presented by Valk et al.

(1990) as well as Bussink and Oenema (1998). Indirect way of lowering nitrogen content in the

urine can be achieved through flushing floors (Braam et al., 1997a), decreasing the PH value at the

emitting surfaces by addition of acid (Monteny and Erisman, 1998).

Swiestra et al. (1995) reported a comparative study on ammonia emission from cubical

houses for cattle with slatted and solid floors; the emission from the compartments with solid floors

was about 50% of the emission of the compartment with the slatted floor. Braam et al. (1997b)

compared two solid floor system (with and without slop) with the traditional slatted floor system;

ammonia emission from the compartment with the non-sloped solid floor was almost equal to that

from a compartment with slatted floor; the sloped solid floor reduced ammonia emission by 21%

compared to the slatted floor. Studies showed that a reduction of ammonia emission can be

expected from design of proper floor systems. But the efficiency was associated with manure

removal systems (Braam et al., 1997a; Braam et al., 1997b). Fast removal of the urine on the floor

may be superior to floor types in terms of ammonia emission reduction.

For a dairy cow building with slatted floors, the airflow pattern and the air speeds near the

slatted floors and in the pit headspace are important factors on ammonia emission (Elzing and

Monteny, 1997). Methodologies on regulating the airflow characteristics in that region can be a

possibility for ammonia abatement. Saha et al. (2010b) applied an extra pit ventilation system to

remove the highly concentrated gases from the pit headspace in a fatting pig room; by utilizing an

air purification unit, reductions in the ammonia emission of 37-53% might be achieved. A similar

research on partial pit ventilation system applied to a naturally ventilated cattle building in reality

has not been found yet.

Page 22: NEERENGI - Aarhus Universitet

8

1.6. Research objectives

The above discussions have highlighted the importance of estimation of ventilation and ammonia

emission rate from naturally ventilated dairy cow buildings. Although much research has been

attempted to solve the issue (Demmers et al., 1998; Zhang et al., 2005; Ngwabie et al., 2009;

Feidler and Müller, 2011; Samer et al., 2012), there is still a gap on techniques to calculate

ventilation rate and determination of sampling positions for representative gas concentration.

Moreover, low-emission technology should be proposed to dairy cattle buildings. A concept of

partial pit ventilation system with air purification units may have the potential to reduce gas

emissions and improve the air quality in cattle buildings. Thus, the main objective of the PhD thesis

was to establish a validated CFD model to predict ventilation rate as well as determine the sampling

positions for representative gas concentration, and to investigate the potential of a partial pit

ventilation system to reduce gas emissions. The specific objectives of this study were to:

measure and quantify ammonia emissions in the two cow buildings and study the effect of

the climatic factors (external wind speed, wind direction, external air temperature) on

ammonia emissions;

evaluate velocity and turbulence characteristics of the airflow in two naturally ventilated dairy

cattle buildings in terms of statistical descriptors, such as autocorrelation, kinetic energy,

turbulence energy dissipation rate, integral time and length scale of turbulence, Kolmogorov

microscale of length;

develop a CFD model to calculate air exchange rate (AER) in one of the cow buildings and

evaluate the performance of tracer gas decay and CO2 balance method for quantifying the

AER as well as determine optimum positions to sample representative CO2 concentration in

the exit air;

investigate the ability of a partial pit ventilation system to reduce gas emissions by scale

model measurements and CFD simulations;

investigate the difference of airflow and pollutant transportation in the pit headspace between

modelling slatted floor directly and modelling slatted floor as porous media using large eddy

simulations;

investigate the potential of a partial ventilation system applied to a full scale naturally

ventilated dairy cattle building to reduce ammonia emissions by adopting CFD method.

1.7. Outline of this thesis

This thesis seeks to determine ventilation and emission rate from naturally ventilated dairy cattle

buildings as well as to propose a systematic approach to curtail gas emissions. In chapter 2,

ammonia and greenhouse gas concentrations were measured in the two buildings. Emissions of

ammonia and methane were quantified using CO2 balance method. The effect of climatic factors

including the external wind speed, direction, the internal airflow velocity, the external and internal

Page 23: NEERENGI - Aarhus Universitet

9

air temperature on emissions of ammonia and methane were discussed. In chapter 3, three

dimensional air velocities were measured outside and inside two naturally ventilated cattle

buildings; the velocity and turbulence characteristics were presented to provide insight into the

variables related to natural ventilation rate and distribution of gas concentrations, consequently

related to gas emissions. Different techniques to estimate ventilation rate was assessed in chapter 4.

The techniques included tracer gas decay, CO2 balance and integration of the volume flow rates

through outlet openings. The sampling locations were discussed to gain representative gas

concentration in the exit air in order to predict gas emissions using CO2 balance method. The

subject of chapter 5 was to test the capability of partial pit ventilation system to reduce emission in

a laboratory condition using a 1:2 scale model. In chapter 6, the similar investigation was also

completed via CFD simulations using Reynolds-averaged Navior-Stokes turbulence models. So far,

partial pit ventilation system was not applied to the existing cattle buildings. Thus the potential of

the system to abate emission was studied by CFD models. The technical issue on dealing with

slatted floors should be addressed before applying partial pit ventilation system to full scale

buildings. The uncertainty of modelling slatted floor as porous media was described in Chapter 7.

The potential of partial pit ventilation system to curtail emissions from full scale buildings was

eventually tested by CFD methods and covered the chapter 8. The final chapter 9 summarized the

main findings of the thesis.

References

Battye, W., Aneja, V., Roelle, P., 2003. Evaluation and Improvement of Ammonia Emissions

Inventories. Atmospheric Environment 37, 3873-3883.

Bjerg, B., Zhang, G., Kai, P., 2008a. CFD investigations of a partly pit ventilation system as method

to reduce ammonia emission from pig production units. The Eighth ASABE International

Livestock Environment Symposium (ILES Ⅷ)

Bjerg B., Zhang, G., Kai, P., 2008b. Porous media as boundary condition for air inlet, slatted floor

and animal occupied zone in numerical simulation of airflow in a pig unit. AgEng2008

International Conference on Agricultural Engineering, Hersonissos, Crete-Greece.

Bjerg, B., Andersen M., 2010. Numerical simulation of a pit exhausts system for reduction of

ammonia emission from a naturally ventilated cattle building. XVIIth world congress of the

international commission of agricultural and biosystems engineering (CIGR). Quebec City,

Canada.

Braam, C. R., Ketelaars, J.J.M.H., Smits, M.C.J., 1997a. Effects of Floor Design and Floor

Cleaning on Ammonia Emission from Cubicle Houses for Dairy Cows. Netherlands Journal

of Agricultural Science 45, 49-64.

Page 24: NEERENGI - Aarhus Universitet

10

Braam, C. R., Smits, M.C.J., Gunnink, H., Swierstra, D., 1997b. Ammonia Emission from a

Double-sloped Solid Floor in a Cubic House for Dairy Cows. Journal of Agricultural

Engineering Research 68(4), 375-386.

Bussink, D.W., Oenema, O., 1998. Ammonia Volatilization from Dairy Farming Systems in Temperate

Areas: A Review. Nutrient Cycling in Agroecosystems 51, 19-33.

Choi, K., Albright, L.D., Timmons, M.B., 1988. An application of the k-ε turbulence model to predict

air distribution in a slot ventilated enclosure. Transactions of the ASAE 33, 274-278.

CIGR, 1984. Report of Working Group on Climatization of Animal Houses – Heat and Moisture

Production at Animal and House Levels. Scottish Farm Building Investigation Unit,

Craibstone, Aberdeen, Scotland.

CIGR, 2002. Report of Working Group on Climatization of Animal Houses – Heat and Moisture

Production at Animal and House Levels. CIGR Section II, Commission International Du

Genie Rural (International Commission of Agricultural Engineering).

Demmers, T.G.M., Burgess, L.R., Short, J.L., Phillips, V.R., Clark, J.A., Wathes., C.M., 1998. First

experiences with methods to measure ammonia emissions from naturally ventilated cattle

buildings in the U.K. Atmospheric Environment 32,285-293.

Demmers, T.G.M., Burgess, L.R., Phillips, V.R., Clark, J.A., Wathes., C.M., 2000. Assessment of

Techniques for Measuring the Ventilation Rate, using an Experimental Building Section.

Journal of Agricultural Engineering Research 76,71-81.

Demmers, T.G.M., Phillips, V.R., Short, J.L., Burgess, L.R., Hoxey, R.P., Wathes., C.M.,

2001.Validation of Ventilation Rate Measurement Methods and the Ammonia Emission from

Naturally Ventilated Dairy and Beef Buildings in the United Kingdom. Journal of

Agricultural Engineering Research 79 (1), 107-116.

Elzing, A., Monteny, G.J., 1997. Ammonia emission from a scale model of a dairy-cow house.

Transactions of the ASAE 40: 713-720.

Feidler, A.M., Müller, H.J., 2011. Emissions of ammonia and methane from a livestock building

natural cross ventilation. Meteorologische Zeitschrift 20 (1), 059-065.

Gyldenkærne, S., Mikkelsen, M.H., 2007. Projection of the Ammonia Emission from Denmark

from 2005 until 2025. Research notes from NERI No.239, 2007.

Hutchings, N. J., Sommer, S. G., Andersen, J.M., W.A.H, 2001. A Detailed Ammonia Emission

Inventory for Denmark. Atmospheric Environment 35, 1959-1968.

Monteny, G.J., Erisman, J.W., 1998. Ammonia Emission from Dairy Cow Buildings: a Review of

Measurement Techniques, Influencing Factors and Possibilities for reduction. Netherlands

Journal of Agricultural Science 46, 225-247.

Nielsen P.V. 1974. Flow in air-conditioned rooms. Ph.D. Thesis, Technical University of

Denmark,23 Copenhagen, Denmark.

Page 25: NEERENGI - Aarhus Universitet

11

Ngwabie, N.M., Jeppsson, K.H., Nimmermark, S., Swensson, C., Gustafsson, G., 2009. Multi-

location Measurements of Greenhouse Gases and Emission rates of Methane and Ammonia

from a Naturally Ventilated Barn for Dairy Cows. Biosystems Engineering 103, 68-77.

Norton, T., Grant, J., Fallon, R., Sun, D.W., 2010a. Assessing the ventilation effectiveness of

naturally ventilated livestock buildings under wind dominated conditions using computational

fluid dynamics. Biosystems engineering 103 (1), 78-99.

Norton, T., Grant, J., Fallon, R., Sun, D.W., 2010b. A computational fluid dynamics study of air

mixing in a naturally ventilated livestock building with different porous eave opening

conditions. Biosystems engineering 106 (2), 125-137.

Pederson, S., Takai, H., Johnsen, J.O., Metz, J.H.M., Groot Koerkamp, P.W.G., Uenk, G.H., Phillips,

V.R., Holden, M.R., Sneath, R.W., Short, J.L.L., White, R.P., Hartung, J., Seedorf, J., Schroder,

M., Linkert, K.H., Wathes, C.M., 1998. A Comparison of Three Balance Methods for

Calculating Ventilation Rates in Livestock Buildings. Journal of Agricultural Engineering

Research 70,25-37.

Pederson, S., 2006. Agricultural Best Management Practices in Denmark. The workshop on

Agricultrual Air Quality: State of Science, 2006. The Ecologial Society of America, 56-67.

Philippe, F.X., Cabaraux, J.F., Nicks, B., 2011. Ammonia Emissions from Pig Houses: Influencing

Factors and Mitigation Techniques. Agriculture, Ecosystems and Environment 141, 245-260.

Samer, M., Fiedler, M., Muller, H.J., Glaser, M., Ammon, C., Berg, W., Sanftleben, P., Brunsch,

R., 2011. Winter Measurements of Air Exchange Rates using Tracer Gas Technique and

Quantification of Gaseous Emissions from a Naturally Ventilated Dairy Barn. Applied

Engineering in Agriculture 27(6), 1015-1025.

Samer, M., Ammon, C., Loebsin, C., Fiedler, M., Berg, W., Sanftleben, P., Brunsch, R.,

2012.Moisture Balance and Tracer Gas Technique for Ventilation Rates Measurement and

Greenhouse Gases and Ammonia Emissions Quantification in Naturally Ventialted Buildings.

Building and environment 50, 10-20.

Saha, C.K., Zhang, G., Ni, J., 2010a. Airflow and Concentration Characterisation and Ammonia

Mass transfer Modelling in Wind Tunnel Studies. Biosystems Engineering 107 (4), 279-287.

Saha, C.K., Zhang, G., Kai, P., Bjerg, B., 2010b. Effects of a partial pit ventilation system on

indoor air quality an ammonia emission from a fattening pig room. Biosystems Engineering

105 (3), 328-340.

Smits, M.C.J., Valk, H., Elzing, A., Keen, A., 1995. Effect of Protein Nutrition on Ammonia

Emission from a Cubicle House for Dairy Cattle. Livestock Production Science 44: 147-156.

Snell, H.G.J., Seipelt, F., Van Den Weghe, H.F.A., 2003. Ventilation Rates and Gaseous Emissions

from Naturally Ventilated Dairy Houses. Biosystems Engineering 86 (1), 67-73.

Page 26: NEERENGI - Aarhus Universitet

12

Strøm, J.S., 1978. Heat Loss from Cattle, Swine and Poultry as Basis for Design of Environmental

Control Systems in Livestock Buildings. SBI-Landbrugsbyggeri 55, Danish Research

Institute, Denmark (In Danish)

Sun, H.W., Keener, H.M., Deng, W., Michel, F., 2004. Development and Validation of 3-D Models

to Simulate Airflow and Ammonia Distribution in a High-rise Hog Building during Summer

and Winter Conditions. Agricultural Engineering International CIGR Journal 6, Manuscript

BC 04 044.

Swiestra, D., Smits, M.C.J., Kroodsma, w., 1995. Ammonia Emission from Cubicle Houses for

Cattle with Slatted and Solid Floors. Journal of Agricultural Research 62, 127-132.

Teye, F.K., Hautala, M., 2007. Measuring Ventilation Rates in Dairy Buildings. International

Journal of Ventilation 6 (3), 247-256.

Valk, H., Klein Poelhuis H.W., Wentink, H.J., 1990. Effect of Fibrous and Starchy Carbohydrates

Inconcentrates as Supplements in a Herbage-based Diet for High Yielding Dairy Cows.

Netherlands Journal of Agricultural Science 38: 475-486.

Van Breemen, N., Burrough, P.A., Velthorst, E.J., Van Dobben, H.F., De Wit, T., Ridder, T.B.,

Reinders, H.F.R., 1982. Soil Acideification from Atmospheric Ammonium Sulphate in Forest

Canopy Throughfall. Nature 299: 548-550.

Zapletal, M., Chroust, P., 2006. Spatial Distribution of Ammonia Emissions on the Territory of the

Czech Republic. The workshop on Agricultrual Air Quality: State of Science, 4 June 2006.

The Ecologial Society of America, 1258-1262.

Zhang, G., Strom, J. S., Li, B., Rom, H.B., Morsing, S., Dahl, P., Wang, C., 2005. Emission of

Ammonia and other Contaminant Gases from Naturally Ventilated Dairy Cattle Buildings.

Biosystems Engineering 92 (3), 355-364.

Page 27: NEERENGI - Aarhus Universitet

13

Chapter 2

Ammonia and methane emissions from two naturally ventilated dairy cattle buildings and the influence of the climatic factors on ammonia emissions

Paper I:

Wu, W., Zhang, G., Kai, P., 2012. Ammonia and Methane Emissions from Two Naturally Ventilated Dairy Cattle Buildings and the Influence of the Climatic Factors on Ammonia Emissions. Atmospheric Environment 61, 232-243.

Page 28: NEERENGI - Aarhus Universitet

14

Abstract

Based on the requirement of the international conventions, there is a pressing need for inventory of NH3, CH4, CO2 and N2O emissions from livestock buildings. The main aim of this study was to quantify the gas emissions and investigate the influence of the climatic factors on emissions. The measurements were carried out in two naturally ventilated dairy cattle buildings with different layouts, floor types and manure management systems during three periods covering winter and summer time. Air temperature and the three dimensional air velocities inside and outside the buildings were recorded over the course of summer period. Emission rates were determined by CO2 production model. The results showed that the internal concentrations of NH3, CH4 and CO2 were increased or decreased simultaneously. Low concentration of N2O was measured outside and inside the buildings; the difference of the concentrations were also very low. The variation of CH4 and CO2 concentrations showed a strong correlation. The NH3 emission rates varied from 32-77 g HPU-1 d-1 in building 1 and varied from 18-30 g HPU-1 d-1 in building 2. The average emission of CH4 was 290 and 230 g HPU-1 d-1from building 1 and 2, respectively. Diurnal pattern was found for NH3 and CH4 emission rates. From multiple linear regression models, there was a significant linear relationship between NH3 emission rates and climatic factors including the external wind speed as well as the air temperature (P<0.001), but not with the external wind directions (P>0.05). Key words: Ammonia emission; air exchange rate; climatic factors; dairy housing; CO2 production model

Page 29: NEERENGI - Aarhus Universitet

15

2.1. Introduction

Emissions of ammonia (NH3) to the atmosphere cause acidification of soil and increase

eutrophication to aquatic systems, while emissions of greenhouse gases (nitrous oxide (N2O),

carbon dioxide (CO2) and methane (CH4)) to the atmosphere influence the global climate.

Approximately 97% of the ammonia emission in Denmark is related to animal husbandry. About

16% of the total greenhouse gas emission originates from the agricultural sector. Denmark has

approved the 1999 protocol to abate the ammonia emission to 56,800 tonnes NH3-N per year by

2010. The Kyoto protocol requires Denmark to reduce the greenhouse gas emission by 22%

compared with the level in 1990 till the first commitment period (2008-2012). Accurate inventories

of NH3 and greenhouse gas emissions from livestock buildings are quite essential to fulfil the

obligations required in the international conventions.

Accurate quantification of gas emissions from livestock buildings to the atmosphere requires

accurate determination of the ventilation rate and the representative gas concentration in the exhaust

air. Natural ventilation rates of livestock buildings can be measured either directly by monitoring

pressure differences over the ventilation openings (Demmers, 1997), or indirectly, using tracer gas

methods (Demmers et al., 1998). Demmers et al. (2001) reported that the pressure difference

method failed to balance the mass flow rates in and out through all openings. The large openings in

naturally ventilated livestock buildings such as dairy cow houses also make the pressure difference

method difficult to be applied. Tracer gas method can be used in two manners: decay and constant

injection. The approach of tracer gas decay cannot be used to monitor ventilation rate continuously.

The method of constant injection of tracer gas is equivalent to the CO2 balance method, which

assumes that CO2 produced by animals can be calculated based on production models (CIGR,

2002). The method can be applied to gain continuous natural ventilation rates and gas emissions.

Therefore, only the CO2 production model is discussed in this paper.

For dairy cattle buildings, ammonia and other gas emission levels primarily depend on

housing system, floor type and manure management system. Several researches have been carried

out in order to reduce emissions by using different floor and manure management systems.

Swierstra et al. (1995) performed measurements of ammonia emissions from an experimental dairy

cattle house with either slatted or solid floors; and concluded that the emission from the

compartments with solid floors and a central gutter was about 50% of the emission from the

compartments with slatted floors. Braam et al. (1997a) investigated ammonia emissions from a

double-sloped solid floor scraped 12 times per day in a mechanically ventilated dairy cow house;

the solid floor with under-floor slurry storage can reduce ammonia by about 50% when compared

with slatted floor. Braam et al. (1997b) compared the ammonia emissions from two different solid

floor systems with emissions from the traditional slatted floor systems; the solid floor without a

slope did not result in significant ammonia reduction while the solid floor with a 3% slope reduced

ammonia emissions by 21% compared to a slatted floor. Zhang et al (2005) estimated emission rates

Page 30: NEERENGI - Aarhus Universitet

16

of NH3, CH4 and N2O from naturally ventilated dairy cattle buildings with different floor types and

manure-handling systems; the lowest ammonia emission was from buildings with solid, drained

floors with smooth surface. However, they also found that urine puddles remained in many

locations in a building with uneven surface of the concrete floor resulted in high ammonia

emissions. Although there are previous studies on quantifying NH3 emissions from different floor

systems, these are limited in numbers and more case studies especially with respect to naturally

ventilated buildings are still needed.

The interaction of the air in the naturally ventilated livestock house and the external

atmosphere causes the dispersion of ammonia and greenhouse gases from livestock buildings to the

surrounding atmospheric environment. The key climatic factors influencing the gas emissions

include the external air temperature, the external wind speeds and directions. Pereira et al. (2010)

reported direct measurements of NH3 from 3 dairy cattle buildings in Portugal and observed

positive relationships (P<0.05) between NH3 emissions and inside mean air temperature by using

multiple linear regression models. Fiedler and Müller (2011) measured the NH3 concentration from

two naturally ventilated cow sheds and found that the derived NH3 emission rate strongly depended

on the outside conditions such as wind direction and speed. Schrade et al. (2012) quantified NH3

emissions from 6 naturally ventilated cubicle barns for dairy cows and employed a linear mixed-

effects model to infer the significant influence of wind speed and outside temperature on NH3

emissions. To map the NH3 emissions with climatic parameters for national inventories of

emissions, there is a strong need for reliable and detailed NH3 emission data as well as climatic

factors at high spatial and temporal resolution.

The objectives of this study were to:

(1) measure ammonia and greenhouse gas concentrations in two naturally ventilated dairy

cow buildings;

(2) quantify the emissions of NH3 and CH4 using CO2 production model;

(3) evaluate the effect of the climatic factors (external wind speed, wind direction, external

air temperature) on NH3 emissions.

2.2. Materials and Methods

The measurement periods in building 1: from Sep-16 to Oct-12, 2010; from Nov-12 to Nov-26,

2010 and from May-31 to Jun-11, 2011. The measurement periods in building 2 were: from Oct-18

to Nov-08, 2010; from Dec-02 to Dec-22, 2010 and from Jun-20 to Jul-14, 2011

2.2.1. Buildings

Two free-stall cubical dairy cattle buildings, which was naturally ventilated, in the Mid-west of

Jutland with different housing and manure management systems were selected for the

investigations. The measurement was not interfered by daily management.

Page 31: NEERENGI - Aarhus Universitet

17

Building 1 (see Fig. 2.1 a) was approximately 85 m long and 24 m wide. Measured from floor

level, the height to the eave was 3.0 m, and the height to the ridge was 7.5 m and the height of the

side wall was 1.2 m. The height of the side wall openings was 1.5 m. The width of the ridge

opening was 0.6 m. Curtains for adjusting sidewall openings were mounted on the low edge of the

opening and can be manually pulled up. The curtain height in the first two measurement periods

was about 0.8 m and the curtains were fully open in the third period. One end of the building was

entirely open with a dimension of 20.11×2.45 m, whereas the other end was closed except for a 5.7

x 2.75 m gate. The milking parlour was located in this end of the building. The feeding alley and

resting area for cow to lie down had a raised platform. A narrow slatted floor in the centre of a

lower walkway for cows was made of concrete. Underneath the slatted floors was manure channel

equipped with scrapers. The manure in the gutters was scraped 12 times a day into manure storage

tanks outside the buildings. Two floor scrapers run 4 times a day to remove the manure on the lower

walkway. There were another two buildings (Fig. 2.1a) near the measured building, which had

dimensions of 58.2 m×24 m×7.6 m and 48.6 m×24 m×8.4 m, respectively. They were 10 m and 49

m away from the measurement building. The closer one was also a dairy cattle house and the farther

one was a storehouse. Behind the measurement building were a workshop, living houses and some

trees in the right corner. Open field were anywhere else.

Page 32: NEERENGI - Aarhus Universitet

18

(a) Building 1

(b) Building 2

Fig. 2.1 Schematic of the measured section in the buildings and the sensor positions: Solid circles marked with A, B, C, D, E, F denote positions for velocity measurements; Thick lines and circles

with cross denote the sampling locations for gas measurements.

Building 1

Building 2

Page 33: NEERENGI - Aarhus Universitet

19

Building 2 (see Fig. 2.1 b) was about 114 m long and 36 m wide. The building was divided

into two sections, in the middle of which was a milking parlour. Measured from floor level, the

height to the eave was 4.3 m, and the height to the ridge was 11.75 m, the height of the side wall

was 1.2 m. The height of the sidewall openings was 1.9 m. There were two gates at one end of the

barn and each of them had a dimension of 4.85×3.40 m. The width of the ridge opening was 2.5 m.

The ridge opening was confined by a ridge space. The detailed roof configuration and the

dimension of the building are shown in Fig.1b. The side wall openings were equipped with

automatically controlled curtains. During winter, the curtains were adjusted according to wind

speed. During summer, they were fully open. The aisle for feeding and the resting area for cow had

a raised platform. The walking alley between the cubicles had slatted floor over a manure channel.

A scraper robot scraped the entire slatted floor every second hour. The pit scrapers under slatted

floor run 4 times a day and remove the manure to an outside slurry tank

2.2.2. Production and feed

In building 1, there were 165 cows with an average weight of 625 kg and an average milk yield of

31 kg per day. The percentage of the milk protein was 3.4%. The daily feed consumption per cow

was: 8 kg grass, 4 kg maize and 8 kg concentrated food. The crude protein was 3.6 kg cow-1 day-1.

Building 2 held 125-128 (Holstein Frisian) dairy cows with an average weight of 625 kg on average

and milk yield of 28-32 kg per day on average. The percentage of the milk protein was 3.7%. There

were also 14-23 calves with a weight of 450-500 kg on average in this section. The feeding

procedure was very complicated due to other experiments. The average feed consumption was

about 47.5 kg cow-1 day-1 with crude protein intake of 3.7 kg cow-1 day-1. The feeding time for the

two buildings was around 11 am. Each cow got individually amount of concentrates in the milking

parlour. The cows were milked twice per day (7 am and 4 pm) in building 1. The cows in building 2

were milked by automatic milking system.

2.2.3. Measurement setup

Gases inside the buildings were sampled by three 20 m long FEP (tetraflouroethylene-

hexaflouropropylene) tubes. Each tube had 20 uniform distributed sampling openings. Each open

was made to ensure the same airflow rate so that the air sampled at each tube represented the

average of the 20 sampling openings. Demmers et al. (2001) reported large error in ventilation rate

estimate using internal sampling points; they also indicated that no obvious zones or locations,

which offered a representative concentration, could be identified within the building section.

Another research carried out by Demmers et al. (1998) showed that the mean tracer gas

concentration measured in the openings was better to represent the gas concentration in the exit air

than that measured around a ring line in the building. Therefore, sampling tubes were placed near

ventilation openings. Two of the tubes (Tube 1 and Tube 3) were at two sides of the building

around 1.2 m away from the side wall openings and 3.0 m above the floor. Tube 2 was below the

Page 34: NEERENGI - Aarhus Universitet

20

ridge opening in the middle of the building and 7.1 m above the floor. Gases at two outside

positions about 2 m from a side wall were also sampled as background reference. The transport

tubes connecting the sampling tubes and the gas monitoring instruments were between 9 m and 50

m long with heating cables attached and were insulated in a plastic foam cover tube to avoid

condensation of water vapor. All the sampling tubes and transport tubes were of FEP

(tetraflouroethylene-hexaflouropropylene) with 6 mm inside diameter and 1 mm of tube thickness.

The gas concentrations were measured applying Innova Photoacoustic Field Gas monitor

1312 coupled with a multipoint sampler 1303 (INNOVA air Tech Instruments A/S, Denmark). The

monitor equipped with filters at the inlet was placed inside a van parked outside the building. Five

channels of the multiplexer 1303 were used to connect to the three indoor sampling lines and two

outdoor measurement positions. Each channel had a suction pump with Teflon membrane (Model

Eg 7130-4AY-RLT, 19w, GEFEG Motoren) with a flow rate of about 30 l s-1 to deliver the sampled

air to the multiplexer. The sampling period for each measurement was 20 s, followed by 20 s

cleaning time to replace the air in the measuring chamber of INNOVA before a new measurement

started. Each channel was repeated six times. The measurement period for each channel was 5 min

before switching to the next channel. To reduce measurement uncertainties caused by delay effects

(Rom and Zhang, 2010), the NH3 concentration at each location was determined by the average of

the last three repeated values in the measurement, while other gas concentrations were calculated as

the mean of all the six repeated values.

Air velocities were measured by three dimensional ultrasonic anemometers – WindMaster

(Gill instruments Ltd, Hampshire, UK). The detailed information of the WindMaster can be found

in Wu et al. (2012). The air velocities inside the buildings were recorded at 6 positions (Fig. 2.1a-b).

Four velocity sensors (C, D, E and F) were located near the sidewall openings. Two velocity

sensors (A and B) were placed in the centre of the buildings near the ridge openings. The positions

for velocity measurement in building 1 and 2 are shown in Fig.1 a-b. The external wind speeds and

directions were also monitored at 10 m above the ground. All the anemometers were connected to a

multi-port adapter supplying electricity power to the anemometers and collecting three-dimensional

velocities. The output frequency was 20 Hz.

The air temperature was recorded at one position inside the building and at one position

outside the building. The temperature sensor was a standard type K thermocouple with an operating

range of -40 to 85 °C and an accuracy of ±1 °C. Temperature readings were taken every 10 min

through the entire measurement period.

2.2.4. Calculations of ventilation and emission rate

The CO2 production model for dairy cows (CIGR, 2002) was used to calculate CO2 produced in

house and used to determine the air exchange rate. Following mass balance, ventilation rate can be

calculated as

Page 35: NEERENGI - Aarhus Universitet

21

outCOinCO

CO

CC

EVR

,,

6

22

210

(2.1)

where, VR was ventilation rate, m3 h-1; 2COE was total CO2 production from the measured building,

kg s-1; inCOC ,2was the CO2 concentration inside buildings, mg m-3; outCOC ,2

was the minimum CO2

concentration of the two outside sampling positions, mg m-3.

The emission rate was expressed as emission per heat production unit (HPU), which was defined as

1000 W total heat produced by the animals at an ambient temperature 20 °C (Zhang et al, 2005).

The emission rate per HPU was thus stated as

totaloutiinii HCCVRE /)( ,, (2.2)

where, Ei was the emission rate per HPU of gas i, mg h-1 HPU-1; i, symbolized the measured gases,

NH3, CH4 and N2O; Ci,in was the average concentration of gas i at all the inside sampling positions,

mg m-3; Ci,out was the minimum CO2 concentration of gas i at the two outside sampling positions,

mg m-3; Htotal was the total HPU produced by cows in the measured building. HPU per cow was

derived based on body weight and milk production without considering pregnancy. The detailed

description of HPU can be found elsewhere (Zhang et al, 2005; CIGR, 2002).

2.2.5. Statistical analysis

A multiple linear regression model was used to describe the relationships between NH3 emission

(ENH3) and the influencing factors such as wind directions (WD), wind speeds (WS), outside air

temperature (TO), inside air temperature (Ti) and inside air velocity (Vi).

Data of ENH3 was preprocessed by a natural logarithmic transformation before the statistical analysis

was carried out. The co-linearity of the influencing factors was also considered to take account into

the effect of the interactions among different variables. This led to the following formula (Verzani,

2004):

)()(log3 jiijiiNHe xxxE (2.3)

where, β was the coefficient of the associated influencing factors; x was the influencing factors,

which had significant effect on ENH3; i and j represented different factors (WD, WS, TO, Ti or Vi );

ji xx denoted the interaction between two influencing factors; ε was the error. The aim of the

multiple linear regression was to comprehensively test the significance (P<0.05) of the influencing

factors on the NH3 emissions.

Page 36: NEERENGI - Aarhus Universitet

22

2.3. Results

2.3.1. Gas concentrations

Daily average concentration of NH3, CH4, CO2 and N2O over the three measurement periods is

given in Fig. 2.2. The internal concentrations of NH3, CH4 and CO2 was increased or decreased

simultaneously.

Fig. 2.2 Daily average concentration of NH3, CO2, N2O and CH4 over the three measurement

periods (solid line-building 1, dashed line-building 2; thick line-inside building, thin -outside

building ). The measurement before and after the breakpoint was finished in 2010 and 2011,

respectively.

Table 2.1 shows the mean value, standard deviation, maximum value and minimum value of

indoor and outdoor gas concentrations of building 1 and 2 during the three measurement periods.

Page 37: NEERENGI - Aarhus Universitet

23

Table 2.1 Indoor and outdoor gas concentrations (Mean, standard deviation, maximum and minimum) of building 1 and 2 during the three measurement periods.

Gas Period

(MM/DD/YYYY) Building

Concentration, ppm Indoor Outdoor

Mean SD* Max Min Mean SD* Max Min

NH3

09/17/2010-10/12/2010 1 1.46 0.79 3.29 0.50 0.34 0.15 0.64 0.0410/18/2010-11/08/2010 2 2.43 0.91 4.31 1.25 0.78 0.11 1.08 0.6311/12/2010-11/26/2010 1 2.54 0.94 4.57 1.46 0.53 0.09 0.76 0.4112/02/2010-12/22/2010 2 3.30 0.89 4.50 1.32 0.73 0.06 0.82 0.5705/31/2011-06/11/2011 1 5.73 1.42 8.62 3.96 0.94 0.18 1.25 0.7306/20/2011-07/14/2011 2 3.03 0.81 4.59 1.90 0.92 0.15 1.15 0.62

CO2

09/17/2010-10/12/2010 1 559 62 697 477 433 11 458 41710/18/2010-11/08/2010 2 701 98 903 564 430 12 466 41411/12/2010-11/26/2010 1 583 82 749 492 422 11 442 40512/02/2010-12/22/2010 2 892 120 1066 598 422 8 439 40805/31/2011-06/11/2011 1 660 54 766 607 452 10 470 43206/20/2011-07/14/2011 2 668 65 805 576 430 12 456 413

CH4

09/17/2010-10/12/2010 1 17.16 5.22 27.48 7.93 6.10 1.76 10.24 3.6510/18/2010-11/08/2010 2 27.65 7.36 42.79 18.13 6.27 2.27 10.21 2.5411/12/2010-11/26/2010 1 17.56 7.15 32.14 9.18 2.73 1.24 5.22 1.2212/02/2010-12/22/2010 2 43.55 10.00 57.33 18.39 1.83 0.35 2.84 1.4805/31/2011-06/11/2011 1 29.88 4.71 39.94 23.44 9.06 1.59 12.83 6.6906/20/2011-07/14/2011 2 27.37 5.18 35.98 19.14 7.85 1.88 12.26 4.50

N2O

09/17/2010-10/12/2010 1 0.33 0.02 0.37 0.29 0.31 0.02 0.35 0.2710/18/2010-11/08/2010 2 0.35 0.03 0.43 0.31 0.34 0.03 0.40 0.3011/12/2010-11/26/2010 1 0.41 0.03 0.47 0.35 0.39 0.03 0.44 0.3312/02/2010-12/22/2010 2 0.47 0.03 0.54 0.38 0.43 0.02 0.45 0.3705/31/2011-06/11/2011 1 0.35 0.02 0.37 0.32 0.33 0.02 0.35 0.3006/20/2011-07/14/2011 2 0.33 0.02 0.36 0.30 0.32 0.02 0.35 0.29

*SD – Standard Deviation

The average concentration of NH3 in building 1 was 1.46, 2.54 and 5.73 ppm for the three

measurement periods, respectively. Average NH3 concentrations of the three periods in building 2

were 2.43, 3.30 and 3.03 ppm. The average CO2 concentrations of three measurement periods were

558, 583, 660 ppm for building 1 and 701, 891, 668 ppm for building 2. The average CH4

concentrations of three measurement periods were 17.16, 17.56, 29.88 ppm for building 1 and

27.65, 43.55, 27.36 ppm for building 2. Inside and outside concentrations of N2O were in the same

level and they were very low. The N2O concentration level was in the range of 0.29-0.47 ppm for

building 1 and 0.30-0.54 ppm for building 2.

Page 38: NEERENGI - Aarhus Universitet

24

A strong correlation (R2 is 0.94 for building 1, 0.97 for building 2) was found between the

concentrations of CH4 and CO2 (Fig. 2.3 a-b). The ratio of CH4 and CO2 concentration was 0.05 for

building 1 and 0.04 for building 2.

(a) Building 1 (b) Building 2 Fig. 2.3 The correlations between the concentration of CO2 and CH4 inside the two buildings

2.3.2. Daily average emission rates of NH3 and CH4 and diurnal variations

The daily averages on NH3 and CH4 emission are shown in Fig. 2.4. The NH3 emission rates varied

Fig. 2.4 Daily average emission rates of NH3 and CH4 over the three measurement periods (Solid

line-building 1 and dashed line- building 2). The measurement before and after the breakpoint was

finished in 2010 and 2011, respectively.

from 32-77 g HPU-1 d-1 in building 1 and varied from 18-30 g HPU-1 d-1 in building 2. NH3

emission was at a relatively high level during both cold and warm periods in building 1, compared

with that in building 2. Another difference of the NH3 emissions from the two buildings was the

extreme high peaks (around 120 g HPU-1 d-1) observed in building 1 but not in building 2. The

Page 39: NEERENGI - Aarhus Universitet

25

variation of NH3 in building 2 was quite smooth. The average emission of CH4 was 290 and 230 g

HPU-1 d-1 from building 1 and 2, respectively.

Examples of the diurnal variations on NH3 and CH4 emission rates are given in Fig. 2.5.

Emission levels in building 1 varied more considerably than in building 2. An apparent diurnal

pattern was found for NH3 and CH4 emission rates in building 1. The highest emissions took place

around noon. The small changes on NH3 (15-45 g HPU-1 d-1) and CH4 (220-300 g HPU-1 d-1)

emission rates in building 2 made the diurnal pattern not that obvious like in building 1. The

emission rates were not significantly different between day and night.

Building 1 Building 2

Fig. 2.5 Examples of diurnal variation of NH3 and CH4 emission rates

2.3.3. The effect of the climatic factors on the air exchange rate

Fig. 2.6 shows the effect of the climatic factors on the ACR. There were no consistent relationships

between ACR and WD for both buildings. According to an analysis using multiple linear

regressions, no significant influence of the wind direction on ACR can be found for both buildings

(P >0.05). Strong correlations (R2=0.80 for building 1, 0.83 for building 2) between ACR and WS

implied that more than 80% of the variance of ACR can be interpreted by the variance of WS.

Building 1 Building 2

Page 40: NEERENGI - Aarhus Universitet

26

Fig. 2.6 The effect of the climatic factors on the air exchange rates of the two buildings.

The coefficients were 6.89 for building 1and 4.14 for building 2. The average of air velocities at the

six measurement positions was used to represent the internal air velocity. The internal air velocity

also showed positive correlations (R2=0.85 for building 1, 0.92 for building 2) with ACR. The

Page 41: NEERENGI - Aarhus Universitet

27

coefficients 37.3 for building 1 and 15.7 for building 2 were 5.4-3.8 times of those for the external

wind speeds. The absolute values of temperature difference between the internal and external air

(|To-Ti|) were mainly below 3 °C. No significant influence of the air temperature difference on ACR

was found for both buildings (P >0.05) in this study.

2.3.4. The effect of climatic factors on NH3 emission rates

The variation of external temperature and NH3 emission rates with respect to time is shown in Fig.

2.7 as an example to illustrate the dependency of NH3 emission rates on the individual climatic

factors. The mean temperature was 17 °C in this period for both buildings. The variations in air

temperature happened more within a day than between days. The temperature variation ranged on

average from 10 to 21 °C for building 1 and 14 to 25 °C for building 2 from the night to the day.

The ammonia emission rate accordingly varied on average in between 36.2 and 152.9 g d-1 HPU-1

for building 1; and in between 20.5 and 48.6 g d-1 HPU-1 for building 2. The occurrence of peaks on

ammonia emission followed the variations in air temperature.

(a)Building 1

(b)Building 2

Fig. 2.7 Variations of external temperature and ammonia emission rates during warmer period for two buildings. Solid line - NH3 emission rate; dashed line – external air temperature.

According to a multiple linear regression, no significant relationship was established between

NH3 emission rate and wind direction (P>0.05) for two buildings. There was a significant effect of

wind speeds on NH3 emission from building 1 (R2=0.72, P<0.001) and from building 2 (R2=0.67,

P<0.001). Outside air temperature was shown to have a significant influence on emission from

Page 42: NEERENGI - Aarhus Universitet

28

building 1 (R2=0.77, P<0.001) and building 2 (R2=0.88, P<0.001). Moreover, the internal air

velocities highly depended on the external wind speeds (R2=0.84 for building 1, 0.85 for building

2). Meanwhile, the internal air temperature perfectly depended on the external air temperature

(R2=0.98 for building 1, 0.99 for building 2).According to the dependence of factor Ti and TO, factor

Vi and WS, only the influence from the independent variables (WD, WS and TO) were considered in

equation 3. The coefficients of equation 3 for NH3 emission rate and the significance of each

variable are given in Table 2.2. The NH3 emission rate predicted by climatic variables was

described as follow.

Building 1: )(040.0211.0828.0)(log3 OONHe TWSTWSE (R2=0.98) (4)

Building 2: )(058.0190.0996.0)(log3 OONHe TWSTWSE (R2=0.99) (5)

Table 2.2 NH3 emission rates (ENH3) as a function of wind speed (WS), external air temperature

(TO), the interaction of WS and TO (WS: TO) Building 1 Building 2

Factors Coefficient SD* P value Factors Coefficient SD* P value WS 0.828 0.039 <0.001 WS 0.996 0.041 <0.001 TO 0.211 0.006 <0.001 TO 0.190 0.003 <0.001

WS×TO -0.040 0.003 <0.001 WS: TO -0.058 0.003 <0.001 *SD-standard deviation

The predicted NH3 emission rate is given in Fig. 2.8. It explained 98% of the total variance of

calculated ammonia emission rate by applying the three independent climatic variables. The

multiple linear regression models also showed that WS (P<0.001) and TO (P<0.001) influenced the

ammonia emission significantly. Moreover, the interaction of WS and TO (WS×TO) had a significant

but negative effect on ammonia emission for both buildings (P<0.001).

2.4. Discussion

2.4.1. Gas concentrations

The transportation and distribution of gas concentrations inside naturally ventilated livestock

buildings depended on air temperature and air velocity (Saha et al., 2010). The results showed that

the difference of NH3, CH4 and CO2 concentrations within both buildings varied following the

temperature and air velocity during different measurement time. The similar results were also found

in several published reports, e.g., Zhang et al. (2005). The low N2O concentrations and low

concentration difference between indoor and outdoor were consistent with the previous study

(Jungbluth, 2001; Ngwabie et al, 2009). Similar N2O indoor concentrations of 0.32-0.40 ppm were

also measured in a dairy cattle building like building 2 with slatted floor (Berges and Crutzen,

1996). Monteny et al. (2006) argued that livestock buildings with liquid manure systems and

external manure tanks were not a major source of N2O and consequently, low emissions of N2O

were found in this kind of buildings.

Page 43: NEERENGI - Aarhus Universitet

29

Building 1 Building 2

Fig. 2.8 Predicted NH3 emission rates by the multiple linear equation versus the measured NH3

emission rates

The linear correlation between the concentration of CH4 and CO2 was consistent with former

research (Ngwabie et al., 2009; Madsen et al., 2010; Bjerg et al., 2011).The ratio of CH4 and CO2

concentration can be used to predict the emission rate of CH4 based on the CO2 production model

(Madsen et al., 2010). The value of the CH4 to CO2 ratio found in this investigation was 0.05 for

building 1 and 0.04 for building 2, while it was 0.08 in the work of Ngwabie et al. (2009). The data

was fitted using a linear regression equation with a negative intercept by Ngwabie et al. (2009).

However, the intercept was set to zero in Fig. 2.3 a-b. This was the reason for the difference of the

ratio in this paper and in the work of Ngwabie et al. (2009). If the data in Fig. 2.3 a-b was fitted

with a negative intercept, the ratio was 0.10 for building 1 and 0.07 for building 2; but the R2

became smaller (0.76 for building 1 and 0.83 for building 2). The ratio was either 0.06 or 0.10

measured by using respiration chamber (Madsen et al., 2010). The ratio in this paper was at the

same level with 0.06

2.4.2. NH3 and CH4 emission rates

The NH3 and CH4 emission rates were determined by CO2 production model (CIGR, 2002). Several

uncertainties related with CO2 balance method may be involved in the calculation of emissions.

CO2 production was based on steady-state body weight and milk yield. This may increase the errors

of emission estimation (Zhang et al., 2005). The accuracy of emission estimation was also

influenced by the gas sampling positions (Demmers et al., 1998; Feidler and Müller, 2011).

However, CO2 production model was the best method for a continuous estimation of gas emissions.

Moreover, CO2 production model used animals as emission sources. Compared with tracer gas

decay method, the randomly distributed animals provided better mix of CO2 and room air.

The NH3 emission rates calculated in this paper were in line with those in literature (Zhang et

al., 2005). Zhang et al. (2005) investigated NH3 emissions in 9 buildings with different floor

Page 44: NEERENGI - Aarhus Universitet

30

systems in Denmark and reported emission rates in the range of 8-76 g HPU-1 d-1; the highest

emission was observed in a building with solid floor and a scraper. The NH3 emission rates in

building 1 with solid floor and scrapers were higher than those in building 2 with slatted floor and

robotic scrapers. One more reason responsible for low emission rates in building 2 could be the

ventilation regulation by automatically adjusted opening curtains. The curtains may reduce both

wind speed and ventilation rates, consequently the emission rate. Low NH3 emission rates of 19-24

g HPU-1 d-1 was also reported in a dairy cattle building with automatically regulated ventilation

flaps mounted on the side walls below the eaves and also at the ridge (Ngwabie et al., 2009).

The average CH4 emission rates of 290 g HPU-1 d-1 for building 1and 230 g HPU-1 d-1 for

building 2 fell in the range of 220-490 g HPU-1 d-1 found in the 9 buildings investigated by Zhang et

al. (2005). In the study of Ngwabie et al. (2009), CH4 emission rates calculated based on HPU were

around 240 g HPU-1 d-1. The CH4 emission rates in this study were near to the enteric CH4 emissions

(234-263 g HPU-1 d-1) measured in a respiratory chamber (Jungbluth et al., 2001). This was

reasonable due to the fact that CH4 in a cow barn was mainly produced from enteric fermentation

(Monteny et al, 2006). The strong correlation between CO2 and CH4 concentrations also indicates

CH4 emission rate may be quickly estimated from the multiplication of CO2 production rate from

cow and a constant coefficient (about 0.06).

The diurnal variations in NH3 and CH4 emissions were also reported by other studies (Zhang

et al., 2005; Ngwabie et al., 2009; Ngwabie et al., 2011). Peaks of gas emissions occurred around or

after noon – one hour after the feeding time. Ngwabie et al. (2011) found that diurnal variations

were related to the feeding schedule, which may change the animal activities. NH3 emissions in

building 2 were less fluctuated than building 1. This could be due to the difference of the floor and

manure removal systems. The robotic scraper in building 2 can move the fresh manure into the pit

in time compared to the scraper operation in building 1. The observed clear floor may keep the NH3

emissions at lower level in building 2.

2.4.3. The effect of the climatic factors

Snell et al. (2003) reported that the wind speed but not direction was shown as a dominant factor of

influencing ACR. During the measurement period, the mean outdoor wind speed was 3.50 m s-1for

building 1 and 2.45 m s-1 for building 2. The airflow inside the two buildings was mainly wind-

driven. Meanwhile, the temperature differences were mainly below 3 °C. In this case, the wind

effect was superior to the stack effect. Therefore, the temperature difference did not show

significant influence on ACR.

Wind speed was previously identified as an important controlling variable for NH3 emission

from the surface of slurry (Sommer et al., 1991) and from the livestock buildings (Snell et al., 2003;

Zhang et al., 2008; Schrade et al., 2012). It was clearly shown that the ventilation rates in the

Page 45: NEERENGI - Aarhus Universitet

31

buildings rose with increasing wind speed (Fig.6), consequently, leading to the increase of NH3

emissions determined by ventilation rates.

The significant and positive linear relationship between emission rate and ambient

temperature in this study was consistent with literature (Misselbrook et al., 2006; Pereira et al.,

2010; von Bobrutzki et al., 2011; Schrade et al., 2012). NH3 formation and release process was

closely associated with temperature. The comparison of the variations of NH3 emission and

temperature in Fig.7 confirmed that higher temperature generally resulted in a considerable

emission increase. The occurrence of peaks on ammonia emission followed the variations in air

temperature. But there was a difference in the trend of variations on temperature and ammonia

emission since temperature was not the only factor that affected the emission. The temperature

gradually increased from night to day, while the overall ammonia emission also increased but with

large fluctuations. The fluctuation could be caused by the rapid fluctuations of wind speeds in the

day or in the night.

In addition to the direct effect of temperature on the NH3 volatilization process, the effect

from the interaction of temperature and wind speed was also significant. The negative coefficient

seemed to imply that the interaction decreased the coherency of each individual factor (either wind

speed or temperature) and the NH3 emission.

Applying F-test to the coefficients in equation (4) and (5), the difference of coefficients for

the associated variables was not significant (P>0.05). For example, the difference of coefficients for

wind speed 0.828 (0.039) in equation (4) and 0.996 (0.041) in equation (5) was statistically

insignificant (P>0.05). This may indicate that gas emission rates from the two naturally ventilated

livestock buildings could be described by the same equation including the associated climatic

factors such as wind speed and ambient temperature. Besides climatic factors, NH3 emission rates

were influenced by many other factors such urea content of tank milk (Schrade et al., 2012) and

animal activity (Ngwabie et al., 2011). To derive NH3 emission rates based on influencing factors,

more inventories on naturally ventilated cattle buildings are still needed and more factors should be

considered

2.5. Conclusions

Ammonia and methane emissions were measured from two naturally ventilated dairy cattle

buildings in Denmark. The following conclusions were drawn.

The internal concentrations of NH3, CH4 and CO2 were increased or decreased simultaneously

due to …. Low concentration of N2O was measured outside and inside the buildings; the difference

of the concentrations were also very low. The variation of CH4 and CO2 concentrations showed a

strong correlation.

The NH3 emission rates varied from 32-77 g HPU-1 d-1 in building 1 and varied from 18-30 g

HPU-1 d-1 in building 2. The average emission of CH4 was 290 and 230 g HPU-1 d-1 from building 1

and 2, respectively. Diurnal pattern was found for NH3 and CH4 emission rates.

Page 46: NEERENGI - Aarhus Universitet

32

From multiple linear regression models, there was a significant linear relationship between

NH3 emission rates and climatic factors including the external wind speed as well as the air

temperature (P<0.001), but not with the external wind directions (P>0.05).

It should be mentioned that the scraper operation frequency in Building1 was considerable

lower than others with similar floor systems. Anyhow it provided some valuable emission

information under such an operation. Future measurements on emissions under different scraper

operation frequencies may benefit to identify the true effects of the scraper frequency.

Estimation of ammonia and other contaminant gas emissions from naturally ventilated

building is still a challenge due to its large uncertainty. Therefore, more field measurements with

well documented information of both airflow characteristics and gas concentration are needed for

accurately quantifying the emission rate and modelling of the emission process. Research on

optimizing gas sampling positions to find more representative gas concentration in the exit air will

be helpful to reduce the uncertainty of estimating gas emissions. Researches may also be focused on

development of simple measurement method to evaluate the effects of different floor and manure

systems on emissions.

References

Berges, M.G.M., Crutzen, P.J., 1996. Estimates of Global N2O Emissions from Cattle, Pig and

Chicken Manure, including a Discussion of CH4 Emissions. Journal of Atmospheric

Chemistry 24(3), 241-269.

Bjerg, B., Zhang, G., Madsen, J., Rom, H.B., 2011. Methane emission from naturally ventilated

livestock buildings can be determined from gas concentration measurements. Environmental

Monitoring and Assessment, DOI 10.1007/s10661-011-2397-8.

von Bobrutzki, K., Müller, H.J., Scherer, D., 2011. Factors Affecting the Ammonia Content in the

Air Surrounding a Broiler Farm. Biosystems Engineering 108, 322-333.

Braam, C. R., Smits, M.C.J., Gunnink, H., Swierstra, D., 1997a. Ammonia Emission from a

Double-sloped Solid Floor in a Cubic House for Dairy Cows. Journal of Agricultural

Engineering Research 68(4), 375-386.

Braam, C. R., Ketelaars, J.J.M.H., Smits, M.C.J., 1997b. Effects of Floor Design and Floor

Cleaning on Ammonia Emission from Cubicle Houses for Dairy Cows. Netherlands Journal

of Agricultural Science 45, 49-64.

CIGR, 2002. Report of Working Group on Climatization of Animal Houses – Heat and Moisture

Production at Animal and House Levels. CIGR Section II, Commission International Du

Genie Rural (International Commission of Agricultural Engineering).

Demmers, T.G.M., 1997. Ventilation of livestock buildings and ammonia emissions. Ph.D. Thesis,

University of Nottingham.

Page 47: NEERENGI - Aarhus Universitet

33

Demmers, T.G.M., Burgess, L.R., Short, J.L., Phillips, V.R., Clark, J.A., Wathes., C.M., 1998. First

experiences with methods to measure ammonia emissions from naturally ventilated cattle

buildings in the U.K. Atmospheric Environment 32,285-293.

Demmers, T.G.M., Phillips, V.R., Short, J.L., Burgess, L.R., Hoxey, R.P., Wathes., C.M.,

2001.Validation of Ventilation Rate Measurement Methods and the Ammonia Emission from

Naturally Ventilated Dairy and Beef Buildings in the United Kingdom. Journal of

Agricultural Engineering Research 79 (1), 107-116.

Feidler, A.M., Müller, H.J., 2011. Emissions of ammonia and methane from a livestock building

natural cross ventilation. Meteorologische Zeitschrift 20 (1), 059-065.

Jungbluth, T., Hartung, E., Brose, G., 2001. Greenhouse Gas Emissions from Animal Houses and

Manure Stores. Nutrient Cycling in Agroecosystems 60(1-3), 133-145.

Madsen, J., Bjerg, B., Hvelplund., T., Weisbjerg, M.R., Lund, P., 2010. Methane and Carbon

Dioxide Ratio in Excreted Air for Quantification of the Methane Production from Ruminants.

Livestock science 129, 223-227.

Misselbrook, T.H., Webb, J., Gilhespy, S.L., 2006. Ammonia Emissions from Outdoor Concrete

Yards Used by Livestock – Quantification and Mitigation. Atmospheric Environment 40,

6752-6763.

Monteny, G., Bannik, A., Chadwick, D., 2006. Greenhouse Gas Abatement Strategies for Animal

Husbandry. Agriculture, Ecosystems & Environment 112 (2-3), 163-170.

Ngwabie, N.M., Jeppsson, K.H., Nimmermark, S., Swensson, C., Gustafsson, G., 2009. Multi-

location Measurements of Greenhouse Gases and Emission rates of Methane and Ammonia

from a Naturally Ventilated Barn for Dairy Cows. Biosystems Engineering 103, 68-77.

Ngwabie, N.M., Jeppsson K.H., Gustafsson, G., Nimmermark, S., 2011. Effects of Animal Activity

and Air Temperature on Methane and Ammonia Emissions from a Naturally Ventilated

Building for Dairy Cows. Atmospheric Environment 45, 6760-6768.

Pereira, J., Misselbrook, T., Chadwick, D.R., Coutinho, J., Trindade, H., 2010. Ammonia Emissions

from Naturally Ventilated Dairy Cattle Buildings and Outdoor Concrete Yards in Portugal.

Atmospheric Environment 44, 3413-3421.

Rom H B; Zhang G. (2010). Time Delay for Aerial Ammonia Concentration Measurements in

Livestock Buildings. Sensors, 10(5), 4634-4642.

Saha, C.K., Zhang, G., Ni, J., 2010. Airflow and concentration characterisation and ammonia mass

transfer modelling in wind tunnel studies. Biosystems Engineering 107 (4), 328-340.

Schrade, S., Zeyer, K., Gygax, L., Emmenegger, L., Hartung, E., Keck, M., 2012. Ammonia

Emissions and Emission Factors of Naturally Ventilated Dairy Housing with Solid Floors and

an Outdoor Exercise Area in Switzerland. Atmospheric Environment 47, 183-194.

Snell, H.G.J., Seipelt, F., Van Den Weghe, H.F.A., 2003. Ventilation Rates and Gaseous Emissions

from Naturally Ventilated Dairy Houses. Biosystems Engineering 86 (1), 67-73.

Page 48: NEERENGI - Aarhus Universitet

34

Sommer, S.G., Olesen, J.E., Christensen, B.T., 1991. Effects of Temperature, Wind speed and Air

Humidity on Ammonia Volatilization from Surface Applied Cattle Slurry. Journal of

Agricultural Science 117, 91-100.

Swiestra, D., Smits, M.C.J., Kroodsma, w., 1995. Ammonia Emission from Cubicle Houses for

Cattle with Slatted and Solid Floors. Journal of Agricultural Research 62, 127-132.

Verzani, J., 2004. Using R for Introductory Statistics. Chapman and Hall/CRC, 1st edition: 432pp.

Wu, W., Kai, P., Zhang, 2012. An assessment of a partial pit ventilation system to reduce emission

under slatted floor-Part 1: Scale Model Study. Computers and Electronics in Agriculture 83,

127-133.

Zhang, G., Strom, J. S., Li, B., Rom, H.B., Morsing, S., Dahl, P., Wang, C., 2005. Emission of

Ammonia and other Contaminant Gases from Naturally Ventilated Dairy Cattle Buildings.

Biosystems Engineering 92 (3), 355-364.

Zhang G; Bjerg B; Strøm JS; Morsing S; Kai P; Tong G; Ravn P, 2008. Emission Effects of Three

different Ventilation Control Strategies - A Scale Model Study. Biosystems Engineering

100(1), 96-104.

Page 49: NEERENGI - Aarhus Universitet

35

Chapter 3

Turbulent air characteristics in two naturally ventilated dairy cattle buildings

Paper II:

Wu, W., Zhang, G., 2012. Turbulent Air Characteristics in Two Naturally Ventilated Dairy Cattle Buildings. (Submitted to a peer review journal).

Page 50: NEERENGI - Aarhus Universitet

36

Abstract

Natural ventilation has been applied to dairy cattle buildings due to its advantages of energy

conservation and low investment. An insight into air velocities and turbulences will be helpful

to improve the microclimate quality in livestock buildings. Air velocity and turbulence were

measured in two naturally ventilated dairy cattle buildings with three dimensional ultrasonic

anemometers. The turbulence was described by autocorrelation, length scales, kinetic energy

and microscales. The results showed autocorrelation coefficients decayed to zero then started

to behave erratically. The distribution of time integral scale was very irregular. The integral

length scale inside the buildings ranged from 1.08 m to 77.57 m. The length scales did not

keep the same level when the external wind conditions were very similar and did not show

positive correlations with the varied external wind speeds. The internal kinetic energy at

different positions had positive correlations with the external wind speed. Larger turbulence

energy dissipation rate (0.49 m2 s-3) was found near side openings, where the integral length

scale was smaller. Kolmogorov microscales were quite isotropy in this study. Air velocities

near downwind side openings possessed higher power spectra, compared with those in the

middle section or near upwind side openings.

Key words: autocorrelation; cross-correlation; spectral analysis; length scales; energy spectrum

Page 51: NEERENGI - Aarhus Universitet

37

3.1. Introduction

Natural ventilation has been applied to dairy cattle buildings due to its advantages of energy

conservation and low investment. However, the microclimate conditions are more complex in

naturally ventilated buildings than mechanically ventilated buildings due to that natural

ventilation highly depends on the external weather. The microclimate conditions in livestock

buildings are of very importance for sustainable livestock production (Ecim-Djuric and

Topisirovic, 2010). One of the microclimate issues is emissions including odour, gas and

particle from naturally ventilated livestock buildings. Intensive research has been done to

investigate gas emissions (Braam et al, 1997; Zhang et al, 2005; Gustafsson et al, 2005;

Ngwabie et al, 2009; Samer et al., 2011). The challenge is the determination of natural

ventilation rate and a representative gas concentration for the exit air. Both the ventilation

rate and the gas concentration distribution depend on the external wind and internal airflow

patterns including air velocities and turbulences. Therefore, knowledge on those specific

parameters is necessary in order to improve the microclimate quality.

Analysis of air turbulence can be achieved by different methods such as calculation of

kinetic energy, dissipation rate, length scale and energy spectrum. Zhang et al (1992)

conducted detailed measurements of air distribution inside a mechanically ventilated room.

They found out a difference among different locations in terms of turbulent kinetic energy

and power spectrums. Heber and Boon (1993) evaluated turbulence characteristics and

airflow patterns generated by a horizontal non-isothermal ventilation air jet in the special case

of a full-scale livestock building for swine production. They reported that the dominant

frequency of air turbulence varied according to locations and turbulence length scales

increased with jet travel distance. These experiments are valuable for understanding the

turbulent characteristics of indoor ventilation flows. However, the experiments were carried

out in mechanically ventilated room spaces. Boulard et al (2000) studied the mean and

turbulence characteristics of air velocity components in an empty greenhouse tunnel by

spectral analysis; analysis of the energy spectra showed that all the locations had similar

spectral levels in the dissipation region. Although the greenhouse tunnel was naturally

ventilated, the correlation between the external wind and the internal air velocities were not

reported.

The objectives of this study are to: (1) evaluate velocity and turbulence characteristics

of the airflow in naturally ventilated dairy cattle buildings in terms of statistical descriptors,

such as autocorrelation, kinetic energy, turbulence energy dissipation rate; (2) analyze the

different scales of turbulence eddies, including integral time and length scale, Kolmogorov

microscale of length; (3) investigate the coherency of the external and internal velocities by

spectral analysis.

Page 52: NEERENGI - Aarhus Universitet

38

3.2. Methods and materials

3.2.1. Experimental details

The air velocity data recorded by Wu et al. (2012a) was used to analyze the characteristics of

the turbulent flow outside and inside the buildings. The measurements were carried out in

two naturally ventilated dairy cattle buildings, denoted as building 1 and building 2 to be

convenient for further description and discussion. Air velocities were measured by ultrasonic

anemometers – WindMaster (Gill instruments Ltd, Hampshire, UK). All the anemometers

were connected to a multi-port adapter with interfaces supplying power to the anemometers

and transferring data to a computer. The frequency of ultrasonic anemometer is crucial. Too

high frequency requires unnecessary instrument speed and it may give the same result with a

comparatively lower but reasonable frequency. However, too low sampling frequency will

result in losing the information of turbulences in small eddies. Larsen (2006) compared the

difference between measurements with 100 Hz and a series of test frequencies; 10 Hz was

found good enough to capture the necessary information on turbulence in naturally ventilated

room. 20 Hz was used in this measurement. More information on the ultrasonic can be found

in the work of Wu et al. (2012b). The measurement periods were May 31th 14:00 to June

16th 10:00 and June 24th 10:39 to July 14th 09:54 in 2011 for building 1 and 2, respectively.

An ultrasonic anemometer was placed 10 m above the ground to monitor the external wind

velocities. Air velocities were measured at 6 positions (Fig.2.1) inside the buildings. Two

positions A, B were in the middle section of the building and near the ridge opening. Two

positions C, D were near one sidewall opening and two positions E, F were near another

sidewall opening. More detailed description of the measurements can be found in the work

of Wu et al. (2012a).

3.2.2. Data analysis

The measurement was conducted for weeks and a large number of data on velocities was

produced. However, it was not possible to analyze all the experimental data. Hence only 16

periods were picked out for data analysis. Table 3.1 gives the measurement time, mean

velocities and standard deviation of the 16 periods categorized as 16 cases. The first 8 cases

had the same external wind direction and air velocity magnitudes at the same level. The last 8

cases had the same wind direction but varied wind speeds.

3.2.3. Background of statistical methods

The appearance of turbulence manifested itself as random fluctuations of the velocity component about a mean value. In other words, velocity exhibits a time dependent behaviour.

The velocity time series ut can be decomposed into a steady mean component tu and a time

fluctuating component 'tu . The decomposition process is known as Reynolds decomposition.

tu over a time period Δt is defined as,

Page 53: NEERENGI - Aarhus Universitet

39

Page 54: NEERENGI - Aarhus Universitet

40

t

tt dtut

u0

1 (3.1)

The time dependent behaviour of the fluctuating component can be described by some

statistical terms, such as variance, correlation and so on.

3.2.3.1. Variance and kinetic energy

The descriptor used to indicate the spread of the fluctuations 'tu about the mean value tu is the

variance 2' )( tu , denoted σ2. The root mean square of the variance expresses the average

magnitude of velocity fluctuations. The variance is of particular importance since it is

proportional to the momentum fluxes induced by turbulence eddies, which cause normal

stresses. Variance has a further interpretation as the half of the mean kinetic energy contained

in the velocity fluctuations of each component. The total kinetic energy of the turbulence at a

given location can be defined as

2

222wvuk

(3.2)

where, u, v and w represent three components of the velocity.

3.2.3.2. Autocorrelation and cross-correlation functions

The structure of the turbulence can be obtained by studying the relationship between the

fluctuations at different times. The autocorrelation function based on time series xt is defined

as

2'2'

''

)()()(''

tt

ttxx

xx

xxR

tt (3.3)

where, τ is the time difference; the autocorrelation function measures the linear dependence

between two points on the same series observed at different time.

Cross-correlation function is based on two velocity time series xt and yt and is defined

as

2'2'

''

)()()(''

tt

ttyx

yx

yxR

tt (3.4)

The two velocity series may be shifted by a certain distance, for example, the

relationship of the turbulence of the air velocities measured at different positions inside the

livestock building can be analyzed by cross-correlation.

Page 55: NEERENGI - Aarhus Universitet

41

The large eddy causes a certain degree of local structure in the flow. Therefore, there

will be correlation between the values of 'tx and a short time later (τ) or at a given location

( 'tx ) and a small distance away ( '

ty ). The correlation will decrease gradually over the time

scale and the space scale of an eddy. In other words, the autocorrelation and cross-correlation

functions measure the time and size scale of a turbulent eddy.

3.2.3.3. The integral time and length scale of turbulence

The integral time and length scale, which represent concrete measures of the average period

or size of a turbulent eddy, can be computed from integrals of the autocorrelation function

with respect to time shift or distance in the direction of the velocity component. The integral

time scale is defined as the integral of autocorrelation over time from zero to the first zero

crossing (Pasquill and Smith, 1983),

n

xxtRt

tt0

)(''

(3.5)

where, )('' nRtt xx

=0; Δt can be perceived as sampling frequency.

The length scale should be calculated following the motion of the air,

txtL (3.6)

3.2.3.4. Turbulence energy dissipation and small scales

Viscous shear stresses perform deformation work which increases the internal energy of the

fluid at the expense of kinetic energy of turbulence. Therefore, turbulence is always

dissipative. The turbulence energy dissipation rate is (Tennekes and Lumley, 1972)

L

ut

3'

(3.7)

The dissipation of turbulence energy mostly takes place at the smallest turbulence

scales-Kolmogorov microscales. The Kolmogorov microscale of length can be formed as

(Tennekes and Lumley, 1972) 4/13 )/( (3.8)

where, is kinetic viscosity, m2 s-1.

3.2.4. Theory of spectral analysis

Velocity time series can also be analysed in the frequency domain by calculating the power

spectrum. This might be useful in identifying where energy is transferred and in quantifying

its dissipation. The spectral density function between two velocity time series is defined as

Page 56: NEERENGI - Aarhus Universitet

42

the Fourier transform of the correlation function, and distributes the variance of wind time

series over frequency. The application of power spectrum is to identify the frequency of the

major energy containing large eddies responsible for dispersion of ventilated air and gaseous

contaminants in a space. The similarity of power spectrum of inside and outside velocities

will help to build the correlation between inside velocity and outside wind. The difference of

them will help to find the mechanism of the effect that building superimposes on inside

velocities. The spectral density is defined as,

h

iw

xxeRwf

tt

2)()( '' (3.9)

where, w is angular frequency (2π times normal frequency).

3.3. Results and discussion

3.3.1. Hourly averaged air velocities outside and inside the buildings

Records of the external wind speeds and the internal air velocities during the entire

measurement period are given in Fig. 3.2 and Fig. 3.3. Outside wind speed and direction

varied significantly over time for both buildings. No sign showed that variations of internal

air velocities followed the change of the external wind speeds. But the internal air velocities

at the positions in the same side showed the same trend of fluctuation. The air velocities

measured at different heights in the middle of the building also showed the same variation

with respect to time. The mean wind speed was 3.61±1.32 m s-1 for building 1 and 2.50±1.31

m s-1 for building 2. The mean velocities at different positions inside the two buildings during

the whole measured period are shown in Table 3.2.

(a) (b)

Fig. 3.2 Measurements of wind direction and speed outside two naturally ventilated buildings (a) Building 1, and (b) Building 2

3.3.2. Sampling period for statistical and spectral analysis

A one-hour period of outside wind speed measurements was chosen outside building 1 and is

shown in Fig. 3.4 a. The sampling frequency was 20 Hz. The mean value of the entire period

was 4.03 m s-1. During this one-hour period, the mean value and standard deviation were

Page 57: NEERENGI - Aarhus Universitet

43

investigated as a function of the time interval over which they were averaged. The mean and

standard deviation reached a nearly constant after 20 minutes (Fig. 3.4 b). A 20-min

measured velocity time series was then used for spectral analysis for each case.

3.3.3. Autocorrelation and integral turbulence length scales

Autocorrelation functions describe the correlation between consecutive values of the time

series. Utilizing the Taylor’s frozen turbulence hypothesis, the autocorrelation of the velocity

fluctuations becomes useful tool for finding repeating patterns such as the presence of a

periodic circulating structure. In this study, autocorrelation analysis was applied to detect

eddies and to calculate integral turbulence time scale.

(a)Position A in building 1

(b)Position B in building 1

c) Position C in building 1

(d)Position D in building 1

(e)Position E in building 1

(f)Position F in building 1

(g)Position A in building 2

(h)Position B in building 2

(i)Position C in building 2

(j)Position D in building 2

(k)Position E in building 2

(l)Position F in building 2 Fig. 3.3 Measurements of air velocities inside two naturally ventilated buildings at

different positions

Page 58: NEERENGI - Aarhus Universitet

44

(a)

(b)

Fig 3.4. Investigation of sampling period: (a) Wind speed over one-hour period; the white line shows one minute average; (b) Mean and standard deviation of the wind speed as a

function of the time period over which the values were averaged.

As an example, autocorrelation functions for outside wind speed and inside velocity

magnitudes at positions A, C, E for case 1 and case 5 are shown in Fig. 3.5. Autocorrelation

coefficients decayed to zero and started to behave periodically, which was consistent with the

work of Brett and Tuller (1991). The decay rate for that at the downwind position E was

faster than other positions in building 1. Whereas, the decay rate at the middle position A was

faster than other positions in Building 2. A determined trend of autocorrelation coefficients

could not be achieved for velocities at different positions. The discrepancy between two

buildings could be caused by the different building dimensions, the different configurations

of the roof opening and the surrounding buildings. According to the study on mechanically

ventilated airflow (Heber and Boon, 1993), autocorrelation coefficients should decay slower

and slower when air travelled from upwind to downwind side in the room. But this was not

the case in the two naturally ventilated buildings. It indicates the airflow patterns inside a

naturally ventilated building were rather complicated due to the numerous openings.

(a) (b)

Fig. 3.5 Auto correlation function for wind velocities at different positions (black line-outside; blue line-C, at upwind side; red line-A, in the middle section of the room; purple

line-E, at downwind side) for case 1 and case 5. (a) building 1 (b) building 2

Page 59: NEERENGI - Aarhus Universitet

45

Table 3.3 shows the integral time scales for the air velocity magnitudes and their three

velocity components. Direction u and v are shown in Fig. 2.1. Direction w was vertical with

respect to ground. The integral time scale determined from summarizing autocorrelation

functions could be used to describe the time required for an eddy to pass the measured

position. The distribution of time integral scale was very irregular.

Table 3.4 shows the integral length scales for the air velocity magnitudes and their three

velocity components. It should be noted that most of the integral length scales for w direction

are of the magnitude level with the dimension of the measurement volume of ultrasonic.

Larger uncertainties were introduced to determine the length scales for w direction. This kind

of uncertainty was also detected by Heber and Boon (1993). The turbulence length scale of

the outside velocity magnitudes ranged from 15.39 m to 120.56 m for all the cases. The

integral length scale inside the buildings ranged from 1.08 m to 77.57 m. The length scales in

building 2 were generally higher compared with those in building 1. For the first 8 cases,

although the external wind condition were very similar, the length scales did not keep the

same level. For case 9-12 and case 13-16, the external wind speed was increased respectively

for building 1 and 2. However, the length scales did not show positive correlations with the

varied external wind speeds. The reason is not clear but could be due to the large difference

of autocorrelations among different time series. The integral time scale and length scale may

be not only influenced by velocity magnitude, building dimensions but also the inherent

structure of turbulence itself.

3.3.4. Turbulence kinetic energy, dissipation rate and small scales

Table 3.5 shows the turbulence kinetic energy for different cases. The average value of the

turbulence kinetic energy for external wind was 1.63 m2 s-2 and was 5 times larger than that

inside the buildings. It became weakest in the middle of the building – 15 to 20 times lower

than that of external wind. Boulard et al (2000) also found a weak kinetic energy in the centre

of a greenhouse tunnel; the kinetic energy was also 10 times lower than that of the external

wind. Kinetic energy increased near the sidewall openings, when compared to the values

measured in the middle of the building. When the external wind was kept at the same level,

the internal kinetic energy at different positions was also at the same level (see in case 1-4

and case 4-8). Table 5 also shows that the kinetic energy has a positive correlation with the

external wind speed (see in case 9-12 and case 13-16). When the external wind speed was

increased, the internal kinetic energy at all the measured internal positions was increased

accordingly.

Table 3.6 shows the turbulence dissipation rates for the air velocity magnitudes and

their three velocity components. The large dissipation rates were mostly found near side

openings, the largest of which was 0.49 m2 s-3 for air velocity magnitude near downwind

opening in building 1. Turbulence energy dissipation rate is the rate of energy supplied from

Page 60: NEERENGI - Aarhus Universitet

46

Page 61: NEERENGI - Aarhus Universitet

47

Page 62: NEERENGI - Aarhus Universitet

48

large eddies to small eddies. Therefore, it should be related with the length scale of the large

eddies. Since the length scales (Table 4) near side openings were generally small, the large

eddies at this region might dissipate into small eddies more quickly. Heber et al (1996) also

reported a trend of larger turbulence dissipation rate where integral length scale was shorter,

although a different formula was used to estimate the dissipation rate. Like turbulence

integral time and length scale, the correlation between dissipation rate and the external wind

speed was not apparent.

Table 3.5 Turbulence kinetic energy of the air velocities for different cases (m2 s-2) Cases External wind A B C D E F

1 1.66 0.07 0.06 0.15 0.14 0.35 0.18

2 1.25 0.05 0.07 0.26 0.21 0.43 0.23

3 1.39 0.05 0.04 0.11 0.13 0.19 0.12

4 1.88 0.07 0.07 0.16 0.15 0.35 0.24

5 1.50 0.13 0.18 0.90 0.75 0.20 0.25

6 1.17 0.04 0.04 0.46 0.54 0.07 0.14

7 1.13 0.09 0.13 0.57 0.49 0.18 0.22

8 0.92 0.09 0.13 0.58 0.58 0.06 0.08

9 0.54 0.03 0.04 0.18 0.12 0.17 0.09

10 1.24 0.04 0.06 0.34 0.29 0.47 0.31

11 1.53 0.06 0.09 0.30 0.26 0.40 0.24

12 2.69 0.10 0.12 0.38 0.35 0.74 0.42

13 1.08 0.09 0.14 0.66 0.59 0.08 0.08

14 1.78 0.04 0.05 0.71 0.83 0.08 0.08

15 2.33 0.12 0.16 1.27 1.27 0.09 0.13

16 4.00 0.17 0.24 2.12 1.97 0.58 1.05

Table 3.7 shows the Kolmogorov mircoscales for the air velocity magnitudes and their

three velocity components. The micro-scale of turbulence ranged between 3×10-5 m near the

sidewall openings to 6×10-4 m for all the positions. The distribution of Kolmogorov

mircoscales were quite uniform at different positions, compared with the values of integral

length scales. They were around 1 mm. The results of small scales in this study was

consistent with the assumption that small eddies could be treated as isotropic eddies

(Tennekes and Lumley, 1972).

3.3.5. Cross-correlation and power spectrum

Analysis of spatial correlation was intended to find the dependent and coherent link between

two velocity time series. The spatial correlation (Table 3.8) between the internal air velocities

and the external wind speeds were generally small. The velocities at upwind side (position C)

showed a little higher correlation with outside wind; the average of the correlation

Page 63: NEERENGI - Aarhus Universitet

49

Page 64: NEERENGI - Aarhus Universitet

50

Page 65: NEERENGI - Aarhus Universitet

51

Table 3.8 Spatial correlations of internal air velocities and external wind at different positions for case 9-16

Cases External wind A B C D E F

Cross-correaltion between external wind and air velocities at different positions

9 1.00 0.04 0.14 0.23 0.26 0.01 0.03

10 1.00 0.08 0.12 0.36 0.30 0.16 0.02

11 1.00 0.02 0.00 0.22 0.20 0.02 0.01

12 1.00 0.01 0.02 0.22 0.15 0.07 0.02

13 1.00 0.19 0.04 0.16 0.19 0.02 0.11

14 1.00 0.16 0.25 0.28 0.26 0.15 0.01

15 1.00 0.05 0.06 0.16 0.12 0.14 0.06

16 1.00 0.15 0.19 0.22 0.22 0.03 0.00

Cross-correaltion between air velocities at location C and those at other positions

9 0.23 0.01 0.09 1.00 0.22 0.11 0.07

10 0.36 0.08 0.09 1.00 0.32 0.25 0.01

11 0.22 0.03 0.02 1.00 0.26 0.10 0.01

12 0.22 0.02 0.01 1.00 0.31 0.13 0.05

13 0.16 0.05 0.12 1.00 0.32 0.11 0.02

14 0.28 0.06 0.13 1.00 0.27 0.18 0.06

15 0.16 0.11 0.04 1.00 0.25 0.17 0.03

16 0.22 0.41 0.43 1.00 0.45 0.18 0.25

coefficients was around 0.15 for two buildings under different wind conditions. Lay and

Bragg (1988) studied the spatial correlations of jet velocity and air velocities at different

separation distances with jet inlet. They reported that the correlation coefficient was below

0.50 after a minimum distance (0.7 m) from jet inlet. In this study, the spatial distance was

more than 13 m among different positions. The correlations of two velocity series could be

interfered by unexpected perturbations. The small spatial correlations indicate that the

dependence between outside wind and inside velocity series cannot be simply described by

cross-correlation.

The cross-correlation among internal air velocities was also investigated. As an

example, the correlation coefficients between air velocities at position C and at A, B, D, E, F

are also shown in Table 3.8. Very low correlation coefficients were also generally found. The

air velocity at position D at the same side with C showed higher correlations (>0.2).

To detect the coherency of the external wind and the internal air velocity, a comparison

of the power spectra of the external wind and the internal velocities at different positions are

given in Fig 3.6 for case 9 to16. External wind contained more energy. The power spectra of

downwind side E possessed more energy than those at upwind side and in the middle section

of the room. For building 1 (case 9-12), the power spectra at high frequency at upwind side

(C) was at the same level as those at E. In the high frequency range, the decay rate of the

power spectra is close to 3/5f , agreeing with that found in the isotropic turbulence (Hinze,

Page 66: NEERENGI - Aarhus Universitet

52

Case 9 Case 10

Case 11 Case 12

Case 13 Case 14

Case 15 Case 16

Fig. 3.6 Spectral energy distribution of the external wind and internal airflow at

position A, C, E (black line-outside; blue line-C, at upwind side; red line-A, in the middle of

the room; purple line-E, at downwind side) for case 9-16Table 3.3 Integral time scale of the

air velocity magnitudes and associated three components for different cases (s)

τ - turbulence time scale; u, v, w – air velocity direction; A, B, C, D, E, F – measuring positions.

Page 67: NEERENGI - Aarhus Universitet

53

1975). The sampling frequency 20 Hz is sufficient to establish the -5/3 slope of the power spectra at high frequencies. 3.4. Conclusions

Air velocity and turbulence were measured in two naturally ventilated dairy cattle buildings

with three dimensional ultrasonic anemometers. Statistical methods and spectral analysis was

applied to study the turbulence characteristics. Following conclusions can be drawn from this

study:

(1) A sampling period of 20 min was enough to reach a constant value for mean and

standard deviation of velocity time series measured in the two naturally ventilated

buildings for a chosen steady state condition;

(2) Autocorrelation coefficients decayed to zero then started to behave erratically. The

distribution of time integral scale was very irregular.

(3) The turbulence length scale of the outside velocity magnitudes ranged from 15.39 m

to 120.56 m for all the cases. The integral length scale inside the buildings ranged

from 1.08 m to 77.57 m. The length scales did not keep the same level when the

external wind conditions were very similar. The length scales did not show positive

correlations with the varied external wind speeds.

(4) When the external wind was kept at the same level, the internal kinetic energy at

different positions was also at the same level. When the external wind speed was

increased, the internal kinetic energy at all the measured internal positions was

increased accordingly.

(5) Larger turbulence energy dissipation rate (0.49 m2s-3) was found near side openings,

where the integral length scale was smaller. Microscale of turbulence ranged 3×10-5 m

near side openings to 6×10-4 m in the middle section. Kolmogorov microscales were

quite isotropy for the natural airflow in this study.

(6) External wind contained more energy. Air velocities near downwind side openings

possessed higher power spectra at both low and high frequency, compared with those

in the middle section or near upwind side openings.

References

Boulard, T., Wang, S., Haxaire, R., 2000. Mean and turbulent air flows and microclimatic

patterns in an empty greenhouse tunnel. Agricultural and Forest Meteorology 100, 169-

181.

Braam, C. R., Smits, M.C.J., Gunnink, H., Swierstra, D., 1997. Ammonia Emission from a

Double-sloped Solid Floor in a Cubic House for Dairy Cows. Journal of Agricultural

Engineering Research 68(4), 375-386.

Brett, A.C., Tuller, S.E., 1991. The autocorrelation of hourly wind speed observations.

Journal of Applied Meteorology 30, 823-833.

Page 68: NEERENGI - Aarhus Universitet

54

Ecim-Djuric, O., Topisirovic, G., 2010. Energy efficiency optimization of combined

ventilation systems in livestock buildings. Energy and buildings 42, 1165-1171.

Gustafsson, G., Jeppsson, K.H., Hultgren, J., Sanno, J.O., 2005. Techniques to Reduce the

Ammonia Release from a Cowshed with Tied Dairy Cattle. Agricultural Engineering

International: the CIGR Journal, 7. Manuscript BC 04 010. Vol. VII. November, 2005.

Heber, A.J., Boon, C.R., 1993. Air velocity characteristics in an experimental livestock

building with non-isothermal jet ventilation. ASHRAE Transactions 99, 1139–1151.

Hinze, J.O., 1975. Turbulence. New York, McGraw-hill.

Larsen, T. S., 2006. Natural ventilation driven by wind and temperature difference. Ph.D.

Thesis. Department of Civil Engineering, Aalborg University.

Lay, R. M., Bragg, G. M., 1988. Distribution of Ventilation air – Measurement and Spectral

Analysis by Microcomputer. Building and Environment, 23 (3), 203-213.

Ngwabie, N.M., Jeppsson, K.H., Nimmermark, S., Swensson, C., Gustafsson, G., 2009.

Multi-location Measurements of Greenhouse Gases and Emission rates of Methane and

Ammonia from a Naturally Ventilated Barn for Dairy Cows. Biosystems Engineering

103, 68-77.

Samer., M., Loebsin, C., Feidler, M., Ammon, C., Berg, W., Sanftleben, P., Brunsch, R.,

2011. Heat balance and tracer gas technique for airflow rates measurement and gaseous

emissions quantification in naturally ventilated livestock buildings. Energy and

buildings 43, 3718-3728.

Tennekes, H., Lumley, J.L., 1972.A first course in turbulence. MIT press.

Wu, W., Zhang, G., Kai, P., 2012a. Ammonia and methane emissions from Two Naturally

Ventilated Dairy Cattle Buildings and the Influence of Climatic Factors on Ammonia

Emission. Manuscript submitted to Atmospheric Environment.

Wu, W., Kai, P., Zhang, 2012b. An assessment of a partial pit ventilation system to reduce

emission under slatted floor-Part 1: Scale Model Study. Computers and Electronics in

Agriculture 83, 127-133.

Zhang, J.S., Christianson, L.L., Wu, G.J., Riskowski, G.L., 1992. Detailed measurements of

room air distribution for evaluating numerical simulation models. ASHRAE

Transactions 98, 58–65.

Zhang, G., Strom, J. S., Li, B., Rom, H.B., Morsing, S., Dahl, P., Wang, C., 2005. Emission

of Ammonia and other Contaminant Gases from Naturally Ventilated Dairy Cattle

Buildings. Biosystems Engineering 92 (3), 355-364..

Page 69: NEERENGI - Aarhus Universitet

55

Chapter 4

Evaluation of methods for determining air exchange rate in a naturally ventilated dairy cattle building with large openings using computational fluid dynamics (CFD)

Paper III:

Wu, W., Zhai, J., Zhang, G., Nielsen, P.V., 2012. Evaluation of methods for determining air

exchange rate in a naturally ventilated dairy cattle building with large openings using

computational fluid dynamics (CFD). (Submitted to a peer review journal).

Page 70: NEERENGI - Aarhus Universitet

56

Abstract

Naturally ventilated dairy cattle buildings are a major source of ammonia and greenhouse

gas emissions. Accurate estimation in gas emissions constitutes the first step towards

reducing the negative impact of emissions on the local environment. The greatest uncertainty

in the emission estimation is the calculation of the air exchange rate (AER). Computational

fluid dynamics (CFD) was used to quantify the AER in a naturally ventilated dairy cattle

building. The objective of this study was to assess the performance of three techniques for

estimating ventilation rates: (1) integration of volume flow rates (VFR) through openings; (2)

tracer gas decay (TGD) and (3) constant tracer gas injection (CTG). In the developed CFD

model, the animal occupied zone (AOZ) was treated as porous media and the resistance

coefficient of porous zone was derived by pressure drops across AOZ using a sub-CFD

model. The results showed that AERs predicted by VFR and TGD were in good agreement

with each other within a large range of wind speeds. The large difference in AER estimation

using CTG and VFR indicates that the mean CO2 concentration of the entire room may not

represent the concentration at the air exit. It may not be suitable to calculate AER using

mean concentration of internal sampling positions. When wind became stronger, the

accuracy of CTG decreased. The gas sampling positions should be located adjacent to the

openings or even in the openings. To reduce the uncertain introduced by wind direction, all

the openings, especially of different azimuths, should possess sampling tubes. The maximum

gas concentrations could be the optimum value to represent the concentration in the exit air.

Key words: porous media; ventilation; tracer gas; emission; sampling position

Page 71: NEERENGI - Aarhus Universitet

57

4.1. Introduction

Naturally ventilated dairy cattle buildings are a major source of ammonia and greenhouse gas

emissions to the atmosphere (Pereira et al., 2010; Ngwabie et al., 2011; Schrade et al., 2012).

Reduction of emissions from dairy cow buildings will contribute to mitigate environmental

pollution. However, it still lacks of reliable measurement and quantification methods of gas

emissions from naturally ventilated livestock buildings. The greatest uncertainty in the

emission estimation is the calculation of natural ventilation rates. Unlike the mechanically

ventilated livestock buildings, it is more difficult to determine the naturally ventilation rates

(Demmers et al., 1998; Demmers et al., 2000; Demmers et al., 2001; Teye and Hautala,

2007). Bruce (1978) applied Bernoulli’s theorem to derive the air exchange rate through a

simple livestock house under rather simple conditions. However, the airflow in a real cattle

building is usually irregular and multidirectional. The large opening may act as an inlet

during one period and as an outlet during another period due to the wind conditions (Zhang et

al, 2005). These facts make it difficult to directly apply the theory in a real dairy cattle

building. To date, numerous methods have been attempted to estimate natural ventilation

rates from livestock buildings. Demmers et al. (2001) measured the pressure differences at

the openings and converted them to air velocities using Bernoulli’s equation; the airflow rates

can be calculated from the air velocities and the opening areas. This method is equivalent to

summing the volume flow rates (VFR) through openings. But the method failed to balance

the mass flow rates through all the openings. The method of tracer gas decay (TGD) has been

successfully used in determining the ventilation rates in livestock buildings (Demmers et al.,

2001; Zhang et al, 2005). The accuracy of TGD method relies on complete mixing of air,

which is unlikely in livestock buildings. CO2 production model has also been demonstrated a

suitable approach by assuming that the animals are the only source of CO2 (CIGR, 2002;

Zhang et al., 2005). The total CO2 production was estimated by cow weight and milk yield

(CIGR, 2002; Zhang et al., 2005). Prior to calculating ventilation rate, the only parameter to

be ascertained in CO2 production model is the background and outlet CO2 concentration. The

continuous CO2 sample in measurement may take place at a few positions due to the

limitation and the capability of the experimental instrument. The average concentration of all

the sampling positions is generally used to represent the outlet CO2 concentration. This

concentration is named as representative CO2 concentration in the exit air. CO2 production

model is equivalent to injecting constant tracer gas and thus is abbreviated to CTG.

In order to accurately quantify the natural ventilation rate, the capability of the three

methods (VFR, TGD and CTG) should be investigated and documented. While the

investigation required detailed information on the internal flow field, an on-farm

measurement may not provide enough velocity and concentration data to compare the above

methods. Given that the detailed airflow inside naturally ventilated livestock buildings can

also be described by Computational Fluid Dynamics (CFD) (Norton et al., 2010a; Norton et

al., 2010b). CFD was applied to determine the air exchange rate (AER) in this paper.

Page 72: NEERENGI - Aarhus Universitet

58

The objectives of this study are to:

(1) develop and validate a CFD model to calculate the AER in a naturally ventilated

dairy cattle building; the CFD model included a sub-model, which was used to

derive the resistance coefficients of the cows to the airflow;

(2) evaluate the performance of TGD and CTG against VFR for quantifying the

AER;

(3) determine optimum positions to sample CO2 to obtain the representative outlet

CO2 concentration.

4.2. Materials and methods

The performance of the three methods namely VFR, TGD and CTG to estimate AER was

evaluated by a CFD model of a naturally ventilated dairy cattle building. Since it was too

comprehensive to include the cows in the CFD model, the animal occupied zone (AOZ) was

treated as a porous volume in order to consider the resistance introduced by cows. The

resistance coefficients of AOZ were obtained by a sub-model. To ensure the accuracy of the

CFD model, thermal boundary conditions for walls and roofs were documented in this work.

The CFD model must be validated by experiments before it was used for quantifying AER.

This section starts with the introduction of a field measurement, which is used for validating

CFD models. The CFD model including the sub-model, mesh and thermal boundary

conditions is given after the description of the measurement. This section ends with a

discussion on how to quantify AER using the three methods in CFD.

4.2.1. The field measurement for model validation

The experimental study of Wu et al. (2012a) was used to validate CFD prediction of natural

ventilation. One of the two naturally ventilated dairy cattle buildings in Wu et al. (2012a) was

employed to establish the CFD model. The dimensions of the full-scale cattle building are

given in Fig. 4.1a. The building had a length of 111.4 m, a width of 36.0 m, an eave height of

4.30 m and a ridge height of 11.7 m. The height of the side wall was 1.20 m. Two sidewall

openings can be controlled by automatically regulating curtains according to indoor thermal

condition and outdoor climate. In the simulated case, the curtain height was 0.82 m that left a

sidewall opening height of 1.08 m. The width of the ridge opening was 0.60 m.

Air velocities and CO2 concentrations inside and outside the building were measured.

Air velocities were measured by ultrasonic anemometers – WindMaster (Gill instruments

Ltd, Hampshire, UK). Detail on WindMaster can be found in the work of Wu et al. (2012b).

All the anemometers were connected to a multi-port adapter, which supplies electricity power

to anemometers and transfers the data to a computer. The three-dimensional air velocities

were measured in a frequency of 20 Hz. An ultrasonic anemometer was placed 10 m above

the ground to monitor the external wind velocities. Air velocities inside the buildings were

Page 73: NEERENGI - Aarhus Universitet

59

recorded at 6 positions (Fig. 4.1b): two were near one sidewall openings; two were placed in

the centre of the buildings near ridge openings; two were near another sidewall openings. The

CO2 concentrations were measured using a Photo-acoustic Multi-gas Monitor, model 1312

and a multiplexer, model 1303 (Innova air Tech Instruments A/S, Denmark). The internal

concentrations were sampled by three 20 m long tubes (Fig. 4.1b) with 20 uniformly

distributed holes. The three tubes were near the three groups of velocity sensors, respectively.

Gases at two outside positions about 2 m from a side wall were also sampled as background

reference. Data measured in October 26th, 2010 was used for validation of CFD simulation.

(a) Geometry

(b) the sensor positions: Solid circles marked with A, B, C, D, E, F denote positions for velocity measurements; Thick lines and circles with cross denote the sampling

locations for gas measurements.

Page 74: NEERENGI - Aarhus Universitet

60

(c) Positions for data analysis (d) The sub-model for AOZ Fig. 4.1 Geometry of the cattle building, the measurement setup and the sub-model for AOZ

4.2.2. CFD Model details

4.2.2.1. The CFD model and the sub-model for AOZ

The CFD model is depicted in Fig. 4.1a. The CFD model was constructed according to the

real dimensions of the full-scale building. Two surrounding buildings were also established in

the model. Since a detailed simulation of animals, slatted floor and partitions within AOZ

will require massive meshes and long time to iterate the calculations, and will cause the

problem of iteration convergence, Sun et al. (2004) and Bjerg et al. (2008) presumed the

AOZ as porous media with a resistance to airflow in order to consider the effect of the

animals on the air motion. The resistance coefficients of porous media needed to be set up in

the simulation but there was not enough data to derive the coefficients by measurement.

Bjerg et al. (2008) assumed a uniform distribution of animals and extracted part of AOZ to

calculate the pressure drop using CFD. The coefficients were obtained by a linear regression

from the pressure drops and the associated inlet velocities. In order to gain the resistance

introduced by the existing of animals, the first step was to develop a sub-model for AOZ to

characterize the animal’s blocking effects on horizontal airflow. The representative geometry

of the sub-model is shown in Fig. 4.1d. The length of the sub-model domain is 17 m; the

width is 4 m; the height is 1.6 m. Similar to a whole AOZ in a building, the sub-model was

divided into a walking area, a feeding area and a laying area. The density of the animals on

the different areas was assumed uniform. Based on the total cow number and the total floor

area, the density of cows in the sub-model was as follow: two cows stood in the feeding area,

one cow lay down in each of the two laying sections, and one cow stood in the walking alley.

The second step was to fit the pressure drops and the associated inlet velocities into the

following relation

uuFDl

p|)|

2

1(

(4.1)

where, l

pis pressure drop per unit length, Pa m-1; D is viscous resistance, m-2; F is inertial

resistance, m-1; is air viscosity, N s m-2; is air density, kg m-3.

Page 75: NEERENGI - Aarhus Universitet

61

4.2.2.2. Numerical method and mesh

The standard k turbulence (Launder and Spalding, 1974) model was used for all the

simulations. SIMPLE scheme was employed to couple pressure and velocity. Second order

upwind method was used to discretize convection term. Residuals, velocity at one point and

mass balance for inlet and outlet are monitored to achieve iteration convergence.

Mesh is the most important factor for an accurate modelling of environmental flow

around and inside buildings. The optimum mesh distribution (Fig. 4.2) and cells number were

determined by conducting a mesh convergence study. Four different grid systems were

constructed for finding a proper mesh. The cells number for each grid system is given in

Table 4.1.

Fig. 4.2 The mesh and boundary conditions

Mesh convergence can be assessed by the level of agreement between velocities

computed using meshes of different resolutions and that using the finest mesh. This

agreement level can be evaluated by root mean square deviation (RMSD):

n

xxxRMSD

n

i ificf

1

2,,1 )(

(4.2)

where, icx , is the velocity from coarse mesh at one position, ifx , is the velocity from the finest

mesh at the same position with icx , . fx is the mean value of ifx , .

The small RMSD value (0.15) (Table 4.1) between grid 3 and grid 4 also indicated that

grid 3 with 1,091,540 cells was optimum in this study.

Page 76: NEERENGI - Aarhus Universitet

62

Table 4.1 Cells number and mesh convergence Mesh grid4 grid3 grid2 grid1

Cells No. 1,997,932 1,091,540 534,802 257,760 Ratio a 1.83 2.04 2.07 -

RMSD b - 0.15 0.34 0.52 a ratio between grid i+1 and grid i (i=1, 2, 3)

b RMSD-Root mean squared deviation between finest mesh grid4 and coarse mesh grid 3, gird 2, grid 1

4.2.2.3. Wind profile

The power law function was proposed to represent the vertical inlet profile of the

computation domain,

)10/(zUu refz (4.3)

C

ukz

2*

(4.4)

)( 0

3*

zz

uz

(4.5)

)/)10ln(( 00

*

zz

Uu ref

(4.6)

where, uz is the vertical mean velocity, m s-1; Uref is the reference velocity at 10 m height, m

s-1; u* is the friction velocity, m s-1; is the von Karman Constant, 0.40-0.42; z is vertical

coordinate of the computational domain, m; z0 is terrain roughness length; kz is the kinetic

energy, m2 s-2; C is a model constant for the k model, 0.09; z is the turbulence energy

dissipation rate, m2 s-3.

As the simulated building located in an open farm land with few trees and buildings,

0z was approximately 0.03 and α was 0.14 (Wieringa, 1992). Riddle et al (2004) observed

significant profile changes along an empty domain when k model was employed without

roughness modification. In a CFD simulation, the flow profiles of the mean speed and

turbulence quantities those were applied at the inlet plane of the computational domain

should be representative of the roughness characteristics of the upstream domain (Blocken et

al, 2007). In CFD software,Fluent, the roughness is expressed in terms of the roughness

height ks and roughness constant Cs (0.5). ks in Fluent is obtained by (Norton et al, 2010a)

Page 77: NEERENGI - Aarhus Universitet

63

ss C

zk 0793.9

(4.7)

ks is about 0.59 m. Ansys Ltd. (2011) mentioned that yp (the distance from centre point P of

the wall-adjacent cell to the wall) should be larger than ks. Accordingly, the size of the wall-

adjacent cell should be larger than1.18 m, which is too coarse for the CFD simulation.

Therefore, ks = 0z proposed by Blocken et al (2007) was used in this simulation.

The simulation for validation was carried out under an external wind speed of 3.00 m s-

1 for U-component and 1.53 m s-1 for V-component at 10 m height above the ground.

Direction U and V are shown in Fig. 4.1a.

4.2.2.4. Thermal boundary conditions

Natural ventilation is driven by wind and thermal buoyancy. To consider the contribution of

thermal buoyancy to ventilation rate, the thermal boundary conditions of the building

envelops and the heat released from animals should be included in the CFD simulation.

Norton et al (2010a) applied the following heat balance equation to estimate the wall

temperature,

0)( rirocicosol QQQQQ(4.8)

where, solQ is solar irradiation, W m-2; coQ , ciQ is convective heat flux with outside and

inside air, W m-2; roQ , riQ is radiant heat flux, W m-2. The net exchange of radiation

between different surfaces were assumed zero. The inner surface and outer surface

temperatures were assumed equal for walls and roofs. The following equation can be used to

calculate roof temperature (Norton et al, 2010b)

skyrcico

skyskyriciocosrf hhh

ThThThIT

,

,

(4.9)

where, s is the fraction of absorbed solar radiation, 0.2 (Kreith and Kreider, 1978); I solar

irradiation, 498.04 W m-3 in this simulation; h is the convective heat transfer coefficient, W

m-2 K-1; T indicates temperature, K; subscript co , ci , sky means convective heat transfer with

outside, inside air and sky. The outside and inside air temperature is obtained from the

measurements, oT =279.2 K, iT =282.2 K. Sky temperature can be derived using (Chen et al,

1995)

4/1)/( STsky (4.10)

Page 78: NEERENGI - Aarhus Universitet

64

where, S is global thermal radiance, 227.375 W m-2; is Stefan-Boltzman constant,

5.6697x10-8.

Convective heat transfer coefficient can be expressed by (Roy et al, 2002)

L

KNuhc

(4.11)

where, K is the thermal conductivity for roof, 0.0242 W m-1 K-1; L is characteristic length,

which is perceived as roof width, 17 m; for turbulent flow, Nusselt number Nu can be given

by (Roy et al, 2002)

3/15/4 PrRe036.0Nu(4.12)

When equation (4.8) was applied to the side walls, the surface temperature will be

w

skyskyrversiciocoso U

ThIThThT ,

(4.13)

skycociw hhhU ; verI is the solar irradiation on the vertical wall, 498 W m-2 in this

simulation.

The thermal boundary conditions are given in table 4.2.

Table 4.2 Thermal boundary conditions Inlet air Ground floor roof side wall

type Temperature, K

Temperature, K

heat flux

Temperature, K

Temperature, K

value 279.2 279.7 0 275.94 276.18

4.2.2.5. Quantifying AER using the three methods – VFR, TGD

and CTG in CFD

VFR method is quite straightforward. The function of the openings can be distinguished by

examination of the flow direction across the associated opening. Then the ventilation rate is

the summation of the volume flow rates across all the inlet or outlet openings. In order to use

TGD, all the computational cells inside the building must be filled with a scalar of initial

fraction – unity. The fraction of the scalar in the computational cells surround the building

must be zero in the initial stage. The CFD simulation is then run to determine the decay rate

of the scalar (Norton et al., 2010a), which is converted to ventilation rate. In CTG method,

CO2 produced by dairy cows was used as tracer gas source. The CO2 production rate was

determined based on the model (CIGR, 2002) by the body weight, milk yield and pregnancy

of the cows and it was released uniformly inside the AOZ. The mean of the CO2

concentration in the entire room was taken as the representative concentration. The

Page 79: NEERENGI - Aarhus Universitet

65

ventilation rate was the multiplication of the CO2 production rate and the reciprocal of the

representative CO2 concentration.

4.3. Results and discussion

In order to display the airflow patterns and profile of velocity, temperature and concentration,

two planes (plane-x and plane-z) and three lines (line 1, line 2 and line 3) are positioned in

Fig. 4.1 c and were taken as the reference positions for data analysis.

4.3.1. Model validation

Pressure distributions in the sub-model (Fig. 4.1 d) with an inlet velocity 0.1 m s-1 are shown

in Fig. 4.3 a. The resistance coefficient by regression (Fig. 4.3 b) was 7.71 m-2 for viscous

force and 0.06 m-1 for inertial force. Bjerg et al (2008) calculated resistance coefficients for

AOZ in a pig production unit; the viscous and inertial resistance coefficient was around 400-

1500 m-2 and 0.4-1.3 m-1, respectively. The lower resistance for the cow AOZ should be

caused by the lower density of cows (12 m2 cow-1) compared with the higher density of pigs

(2.36 m2 cow-1).

(a)

(b)

Fig. 4.3 Results from the sub-model: (a) Pressure drop (Pa) with an inlet velocity 0.1 m s-1;

(b) the relationship between pressure drop and inlet air velocity. Circle denotes the simulated

values from the sub-model; line denotes the values from the equation 1.

4.3.1.1. Airflow patterns and validating air velocities

Horizontal and vertical airflow patterns at plane-z are shown in Fig. 4.4. The downwind side

opening was divided into left and right section by a wall (the purple line).Two big vortices

were formed and maintained in the right section. The upwind opening acted as air inlet; the

ridge and downwind openings mainly acted as air outlet in this case.

Page 80: NEERENGI - Aarhus Universitet

66

Fig. 4.4 Horizontal and vertical airflow patterns (Purple lines mark the side walls)

Table 4.3 shows the measured and simulated air velocities at different positions.

Overall, good agreement was generally found between the simulated and measured air

velocities. The simulated velocities in the middle section of the room were within the

variation of the measured fluctuated data. For example, at the position near ridge opening, the

simulated velocity was 0.22 m s-1, the measured velocity ranged from 0.10 to 0.30 m s-1

considering the mean velocity 0.20 m s-1 and its fluctuation 0.10 m s-1. A relatively larger

discrepancy existed at the position about 10 m away from the downwind side wall, where the

simulated and measured velocity magnitude was 0.52 and 0.29±0.08 m s-1, respectively. Near

this position, the direction of the outward flow through the opening was altered by the

separated flow that began at the leading edge of the neighbour building (Fig. 4.4). In such a

region the standard k model may have challenge. It was reported that standard k

model has difficulty to predict the flow separation (Castro and Apsley, 1997; Norton et al,

2010b). More uncertainties would be enrolled in the calculation of velocities near a

separation region.

Table 4.3 Velocity magnitude of measurement and simulation Coordinate in domain

Position Measurement

(m s-1) Simulation

X Y Z (m s-1) 19.5 1.75 2.85 Near side opening 0.25±0.09 0.43 31.2 1.75 2.85 Near side opening 0.29±0.08 0.57 27.3 18 7.8 In the middle 0.16±0.09 0.18 27.3 18 9.2 Near ridge opening 0.20±0.10 0.22

4.3.1.2. Temperature distribution

The mean temperature (Fig. 4.5) from CFD simulation was around 281 K in the room. High

temperature 284.7 K was found in the downwind side of the AOZ. The right section of the

room for calves had a uniform temperature around 280.6 K. The vertical temperature

distribution showed that heat was brought up from the AOZ to the roof and the ridge opening

functioned to discharge the heat.

Page 81: NEERENGI - Aarhus Universitet

67

Fig. 4.5 Horizontal and vertical temperature distribution (K)

Cooper et al. (1998) developed a thermal balance model to estimate the mean internal

temperature of livestock buildings. A mean temperature 282.2 K for inside room was

estimated using this model. As described early, the solar radiation and global thermal

radiance were not simulated directly but were transformed into thermal conditions of roofs

and sidewalls using equation (4.8). The heat production from animals was estimated using a

heat production model (CIGR 2002). The mean room temperature 281 K calculated by CFD

agreed with the mean room temperature 282.2 K predicted by the thermal balance model of

Cooper et al (1998).

4.3.1.3. Validating CO2 concentration and AER

The horizontal distribution (Fig. 4.6) showed that high CO2 concentration was found near the

flow vortex. From the vertical field, most CO2 was confined inside the AOZ and a part of it

was transported to upwind side below the roof due to the air recirculation in the upwind side

(Fig. 4.4).

Fig. 4.6 Horizontal and vertical CO2 concentration distribution (mg m-3)

Both simulated and measured CO2 concentrations are presented in table 4.4. Very good

agreements were found. The mean concentration of three positions was 175.9 mg m-3 for

measurement and 167.5 mg m-3 for simulation. The relative difference was smaller than 5%.

Page 82: NEERENGI - Aarhus Universitet

68

Table 4.4 CO2 concentrations and AER from measurement and simulation

Position CO2 (mg m-3) AER (h-1)

Simulation Measurement Simulation Measurement Near the upwind side opening 18.5 23.2 103.6 82.6

In middle of the room 162.3 149.5 11.8 12.8

Near the downwind side opening 321.7 355.0 6.0 5.4

Mean of three positions 167.5 175.9 11.4 10.9

Table 4.4 also shows the AER calculated using simulated and measured CO2

concentration and the variability of AER with respect to CO2 sampling positions. The

representative CO2 concentration was the mean of those at the three sampling positions. The

measured AER 10.9 h-1 was consistent with simulated AER 11.4 h-1. The relative difference

between them was 4.6%.

Through a thoughtful validation of velocity, temperature and concentration, the CFD

model was proved to be accurate and can be used for further systematic investigation and

analysis.

4.3.2. Assessing the capability of VFR, TGD and CTG to calculate AER

To study the performances of the three techniques– VFR, TGD and CTG, AER was

calculated at different wind speeds ranging from 2.83 to 12.73 m s-1 with the same wind

direction of 315°. The predicted AERs are showed in Fig. 4.7. AER predicted by TGD and

VFR had a small deviation. The AERs calculated by CTG using the mean concentration of

the entire room were consistently higher than those by VFR and TGD methods for all the

cases. For example, the simulated AER calculated by VFR and TGD was 69.9 and 68.5 h-1

while the AER calculated by CTG was 106.3 h-1, when the external wind speed was 12.73 m

s-1. The relative difference between the results from CTG and VFR was 50%, which was very

large. The discrepancies can be explained by the requirements of using CTG. The accuracy of

CTG highly depended on the selection of the representative CO2 concentration at the air exit.

The average concentration of the entire room used for calculation of AER may not represent

the outlet gas concentration. In this case, the average concentration of the entire room was

smaller than the truly representative concentration. Fiedler and Müller (2011) indicated that

more sampling positions were required so that the mean concentration of all sampling

positions can be closer to the representative value. However, this study implies that even the

mean concentration of the entire room may not be an appropriately representative

concentration at the air exit. Therefore, using mean concentration of internal sampling

positions would add more uncertainties to the estimation of ventilation. Demmers et al.

(2000) also reported large error in ventilation rate estimate using internal sampling points;

Page 83: NEERENGI - Aarhus Universitet

69

they also indicated that no obvious zones or locations, which offered a representative

concentration, could be identified within the building section.

Fig. 4.7 Air exchange rate (h-1) calculated by three techniques using CFD. Square – VFR;

Circle – TGD; Triangle – CTG. The upper line – fitted using CTG data; the lower line – fitted

using VFR data.

Among the three techniques used in CFD, VFR was the direct way to obtain ventilation

rates and was thus set as the reference method in this study. Compared with VFR, TGD also

highlighted an accurate measure of ventilation rates. The result was consistent with the work

of Norton et al. (2010a). However, TGD is time consuming in CFD simulations. It requires

the calculation in a transient mode, and data at different time steps need to be saved and

handled individually. Although CTG can be easily implemented in CFD simulations, its

accuracy (Fig. 4.7) is questionable due to the fact that a representative concentration cannot

be easily identified if the sampling positions are located inside the livestock building.

Fig. 4.7 shows a strong correlation (R2=1) between wind speed and AER based on the

methods – VFR and CTG. High wind speed causes high ventilation rates. When wind speed

varied from 2.83 to 12.73 m s-1, AER increased from 15.5 to 69.8 h-1 predicted by VFR and

from 24.5 to 106.3 h-1 predicted by CTG. The same correlation between AER and wind speed

was also found in the work of Wu et al. (2012a). The deviation of AER obtained by CTG and

VFR increased from 8.9 to 36.4 h-1. The reason is that stronger wind leads to poorer mixing

of indoor air (Norton et al., 2010b; Teye and Hautala, 2007). The poorer air mixing makes the

concentration of the entire room farther away from the truly representative concentration at

the air exit. Hence, the accuracy of such a CTG method decreases when the wind becomes

stronger.

4.3.3. The representative CO2 concentration at the exit air

The mean gas concentration of the entire room failed to represent that in the exit air for

calculating AER using CO2 production model. In order to find the optimum representative

concentration, AERs were calculated based on CTG using the gas concentrations at different

positions (Tube 1, Tube 2, Tube 3, see Fig. 1b). For the convenience of further analysis, the

Page 84: NEERENGI - Aarhus Universitet

70

CTG method using the mean concentration of the entire room, the mean concentration of the

three tubes, the mean concentration of the tube (Tube 2) near ridge opening and the

downwind sampling tube (Tube 1), the mean concentration of the downwind sampling tube

(Tube 1) was abbreviated to CTG, CTG1, CTG2, CTG3, respectively, without further claim.

Fig. 4.8 a shows the AERs determined with different representative CO2 concentrations

under different wind speeds, when the wind blew to the building with an angle of 315°.

(a) (b)

Fig. 4.8 AERs calculated with CTG method using different representative CO2 concentrations

in the exit air: circle – using mean concentration of the entire room; square – using mean

concentration of the three sampling positions (Fig.1 b); triangle- using mean concentration of

the middle and downwind sampling positions; diamond – using mean concentration of the

downwind position. ACRs calculated with VFR (solid circle) were also plotted as reference.

(a) shows AERs with respect to different wind speeds; (b) shows AERs with respect to

different wind directions.

A large difference was found in AERs predicted using the mean concentration at

different sampling positions. It reveals the high sensitivity of CTG method according to the

sampling locations. AERs predicted by CTG3 and VFR were consistent within a large range

of wind speeds (2.83 - 12.73 m s-1). For example, the AER was 14.7 h-1 based on CTG3,

which had a relative difference of 5.2% with 15.5 h-1 calculated by VFR, when the wind

speed was 2.83 m s-1. While the wind speed varied to 12.73 m s-1, the relative difference

between 64.2 h-1 based on CTG3 and 69.9 h-1 based on VFR was 8.2%. AERs were

overestimated by CTG1 compared to those by VFR and the overestimation became larger

when the wind speed increased. AER calculated by CTG1 ranged from 28.2 to 123.1 h-1 when

the wind speed varied from 2.83 - 12.73 m s-1. The largest relative difference with that

estimated by VFR was 76%. Tube 3 in the simulations was located near the upwind opening.

The concentration sampled by this tube was close to that in the background. The mean

concentration of the three tubes may underestimate the representative gas concentration by

taking the mean concentration of the upwind tube into the average. Consequently, the AERs

were overestimated by CTG1. CTG2 showed a better performance to estimate AER

Page 85: NEERENGI - Aarhus Universitet

71

compared with CTG and CTG1. However, CTG2 also overestimated AERs significantly

during different wind speeds. Demmers et al. (1998) estimated ventilation rates in a dairy

cattle building based on CTG using two sampling systems; the tracer concentration was

measured at nine positions in openings around the perimeter of the building (CTG3) as well

as around a ring sampling line in the building (CTG); they reported that during windy

weather (>5 m s-1) the ventilation was overestimated by the ring line (CTG) compared to the

perimeter samples (CTG3). The results in this work were consistent with that of Demmers et

al. (1998).

Fig. 4.8b shows the AERs determined by different representative CO2 concentrations

under different wind directions with the same wind speed of 4 m s-1. The wind direction was

of outstanding effect on estimation of AER using different representative CO2 concentrations.

When wind direction was smaller than 225°, all the methods based on constant tracer gas

(CTG, CTG1, CTG2 and CTG3) failed to predict AERs at the same level with VFR. The

reason is that the CO2 concentration used in the three methods cannot represent the outlet

concentration for these cases according to the external wind direction. The estimation on

AER ranged from 4.5 to 14.8 h-1 by VFR and from 28.2 to 51.3 h-1 by CTG. The

overestimation by CTG was intolerable. AERs were more overestimated by CTG1, CTG2

and CTG3 compared to those by CTG. When the wind direction became larger than 225°,

CTG3 started to show a better capability to estimate AERs compared with VFR. Therefore,

the mean concentration of Tube 3 can represent the concentration in the exit air but subject to

the wind directions. It can be explained by the function as inlet and outlet of the openings

through examining the airflow patterns at different wind directions. Fig. 4.9 shows the

airflow patterns coloured with CO2 concentration distribution on plane-x. When the wind

blew parallel to the openings (180°), most parts of the long opening on the upper part of the

graph (Fig. 4.9a) and the right opening on the lower part of the graph (Fig. 4.9a) acted as air

inlet; the left opening on the lower part of the graph (Fig. 4.9a) acted as air outlet; that can

explain why the CTG3 did not show a better performance compared with CTG. When the

wind direction varied to 195°, most parts of the three sidewall openings acted as air inlet (Fig.

4.9b); the mean CO2 concentration near sidewall openings thus cannot represent the

concentration in the exit air. When the wind direction became larger than 225°, the long

opening on the upper part of the graph (Fig. 4.9c) acted as air inlet and the two openings on

the lower part of the graph (Fig. 4.9c) started to act as air outlet; consequently, the mean

concentration of the Tube 3, which was near downwind openings, can well represent the

concentration in the exit air; therefore, under this circumstance, CTG3 estimated more

accurate AERs compared with VFR method.

Through the above discussions, it can be concluded that the gas sampling positions

should be close to the openings or even in the openings; the maximum gas concentrations in

the different openings could be the optimum value to represent the concentration in the exit

Page 86: NEERENGI - Aarhus Universitet

72

air. The sampling positions located inside the room may introduce more errors in estimation

of AER.

(a)180° (b)195°

(c) 225° (d)270°

Fig. 4.9 Airflow patterns coloured with CO2 concentration distribution at different wind directions

4.3.4. Further discussion

Based on the requirement of the international conventions, there is a pressing need for

inventory of NH3, CH4, CO2 and N2O emissions from naturally ventilated livestock buildings

(Wu et al., 2012a). The remained and greatest uncertainty in the estimation of the above gas

emissions was the calculation of the ventilation rates. The main purpose of this article was to

evaluate the performance of the different and existing methods to determine ventilation rates.

The above-mentioned performance of the three techniques does not match their applicability

while they are used in field measurements. Naturally ventilated livestock buildings generally

possess large openings which make the velocity measurement in the openings rather difficult.

Therefore, VFR seems impossible to be implemented in reality. Demmers et al. (1998)

indirectly used VFR via measuring the pressure difference across the openings. But the

method failed to balance the mass flow rates in and out through the buildings. When TGD is

applied to an on-site building, the building must be fully covered initially in order to

distribute the tracer gas uniformly across the entire building. This is perhaps not suitable

when the animals are inside the building. Since ventilation rates depend on the external wind

conditions, it will be very difficult to measure the ventilation rates continuously by TGD

method. These are why the CTG method is very popular in the field measurement. CTG is

Page 87: NEERENGI - Aarhus Universitet

73

usually employed in form of an equivalent approach – CO2 production model (Zhang et al.,

2005; Feidler and Müller, 2011).

When the CO2 production model is applied to estimate natural ventilation rates, the key

problem is to ascertain the outlet CO2 concentration since the total CO2 production and

background CO2 concentration can be easily determined or measured. It can be concluded

from this study that the gas sampling positions should be located adjacent to the openings or

even in the openings. To reduce the uncertain introduced by wind direction, all the openings

especially of different azimuths should possess sampling tubes. The maximum gas

concentrations in the different openings could be the optimum value to represent the

concentration in the exit air.

This work can be helpful to design the field measurements of gas emissions from

naturally ventilated livestock buildings; especially can guide engineers to determine the gas

sampling positions when CO2 production model is applied to estimate natural ventilation

rates and emission rates.

The AOZ was treated as porous media in this paper. The uncertainty was not presented

due to the lack of data on air velocities and gas concentrations in the AOZ. The advantages of

using porous media were: (1) simplifying the geometry, consequently, reducing the grid cells

and simplifying the boundary conditions; (2) generally, the cows were randomly distributed

and randomly moved in AOZ; thus, a prescribed uniform CO2 release in porous media was

also reasonable. The limitation of using porous media was that the derivation of airflow

resistance coefficients was not validated against experimental data. In addition, the difference

of using porous media and simulating the cows directly was unknown. Such a difference can

be future work.

4.4. Conclusion

The performances of three techniques – integrating volume flow rates (VFR), tracer gas

decay (TGD) and constant tracer gas (CTG) were assessed by CFD to quantify ventilation

rates from a naturally ventilated livestock building. The following conclusions can be drawn:

(1) The developed CFD model documented with detailed thermal boundary conditions

was validated by air velocities and CO2 concentration data; the resistance coefficients

of AOZ were derived by a sub-model; the coefficient was 7.71 m-2 for viscous force

and 0.06 m-1 for inertial force.

(2) AERs predicted by VFR and TGD were in good agreement with each other within a

large range of wind speeds.

(3) Large difference in AER estimation using CTG and VFR indicates that the mean CO2

concentration of the entire room may not represent the outlet concentration.

(4) When wind became stronger, the accuracy of CTG decreased.

The gas sampling positions should be located adjacent to the openings or even in the

openings. To reduce the uncertain introduced by wind direction, all the openings especially of

Page 88: NEERENGI - Aarhus Universitet

74

different azimuths should possess sampling tubes. The maximum gas concentrations in the

different openings could be the optimum value to represent the concentration in the exit air.

References

Anonymous, 2009. Fluent help Release 12.0. Ansys Inc., Southpointe, 275 Technology

Drive, Canonsburg, PA 15317, USA.

Bjerg B., Zhang, G., Kai, P., 2008. Porous media as boundary condition for air inlet, slatted

floor and animal occupied zone in numerical simulation of airflow in a pig unit.

AgEng2008 International Conference on Agricultural Engineering, Hersonissos, Crete-

Greece.

Blocken, B., Stathopoulos, T., Carmeliet, J., 2007.CFD simulation of the atmospheric

boundary layer: wall function problems. Atmospheric environment 41, 238-252.

Castro, I.P., Apsley, D.D., 1997. Flow and dispersion over topology: a comparison between

numerical and laboratory data for two dimensional flows. Atmospheric Environment

31, 839-850.

Chen, B., Clark, D., Mei, W., Kasher, J., 1995. Measurement of Night Sky Emissivity in

Determining Radiant Cooling from Cool Storage Roofs and Roof Ponds. Proceedings

of the American Solar Energy Society Annual Meeting, Minneapolis, MN.

CIGR, 2002. Report of Working Group on Climatization of Animal Houses – Heat and

Moisture Production at Animal and House Levels. CIGR Section II, Commission

International Du Genie Rural (International Commission of Agricultural Engineering).

Cooper, K., Parsons, D.J., Demmers, T., 1998. A thermal balance model for livestock

buildings for use in climate change studies. J. Agric. Eng. Res. 69, 43-52.

Demmers, T.G.M., Burgess, L.R., Short, J.L., Phillips, V.R., Clark, J.A., Wathes., C.M.,

1998. First experiences with methods to measure ammonia emissions from naturally

ventilated cattle buildings in the U.K. Atmospheric Environment 32,285-293.

Demmers, T.G.M., Burgess, L.R., Phillips, V.R., Clark, J.A., Wathes., C.M., 2000.

Assessment of Techniques for Measuring the Ventilation Rate, using an Experimental

Building Section. Journal of Agricultural Engineering Research 76, 71-81.

Demmers, T.G.M., Phillips, V.R., Short, J.L., Burgess, L.R., Hoxey, R.P., Wathes., C.M.,

2001.Validation of Ventilation Rate Measurement Methods and the Ammonia Emission

from Naturally Ventilated Dairy and Beef Buildings in the United Kingdom. Journal of

Agricultural Engineering Research 79 (1), 107-116.

Feidler, A.M., Müller, H.J., 2011. Emissions of ammonia and methane from a livestock

building natural cross ventilation. Meteorologische Zeitschrift 20 (1), 059-065.

Kreith, F., Kreider, J.F., Principles of Solar Energy, Hemisphere Publishing, New York,

1978.

Page 89: NEERENGI - Aarhus Universitet

75

Launder, B.E., Spalding, D.B., 1974. The numerical computational of turbulent flows.

Computer Methods in Applied Mechanics and Engineering 3, 269-289.

Ngwabie, N.M., Jeppsson, K.H., C., Gustafsson, G., Nimmermark, S., Swensson, 2009.

Effects of Animal Activity and Air Temperature on Methane and Ammonia Emissions

from a Naturally Ventilated Building for Dairy Cows. Atmospheric Environment 45,

6760-6768.

Norton, T., Grant, J., Fallon, R., Sun, D.W., 2010a. Assessing the ventilation effectiveness of

naturally ventilated livestock buildings under wind dominated conditions using

computational fluid dynamics. Biosystems engineering 103 (1), 78-99.

Norton, T., Grant, J., Fallon, R., Sun, D.W., 2010b. A computational fluid dynamics study of

air mixing in a naturally ventilated livestock building with different porous eave

opening conditions. Biosystems engineering 106 (2), 125-137.

Norris, S.E., Richards, P.J., 2010. Appropriate Boundary Conditions for Computational Wind

Engineering Models Revisited. The fifth international Symposium on computational

wind engineering (CWE2010), Chapel Hill, North California, USA.

Pereira, J., Misselbrook, T., Chadwick, D.R., Coutinho, J., Trindade, H., 2010. Ammonia

Emissions from Naturally Ventilated Dairy Cattle Buildings and Outdoor Concrete

Yards in Portugal. Atmospheric Environment 44, 3413-3421.

Riddle, A., Carruthers, D., Sharpe, A., McHugh, C., Stocher, J., 2004. Comparisons between

FLUENT and ADMSfor Atmospheric Dispersion Modelling. Atmospheric

Environment 38 (7), 1029-1038.

Roy, J.C., Boulard, T., Kittas, C., Wang, S., 2002. Convective and Ventilation Transfers in

Greenhouses, Part 1: the Greenhouse considered as a Perfectly Stirred Tank.

Biosystems engineering 83 (1), 1-20.

Schrade, S., Zeyer, K., Gygax, L., Emmenegger, L., Hartung, E., Keck, M., 2012. Ammonia

Emissions and Emission Factors of Naturally Ventilated Dairy Housing with Solid

Floors and an Outdoor Exercise Area in Switzerland. Atmospheric Environment 47,

183-194.

Sherman, M.H., 1990. Tracer-gas Techniques for Measuring Ventilation in a Single Zone.

Building and Environment 25 (4), 365-374.

Snell, H.G.J., Seipelt, F., Van Den Weghe, H.F.A., 2003. Ventilation Rates and Gaseous

Emissions from Naturally Ventilated Dairy Houses. Biosystems Engineering 86 (1), 67-

73.

Sun, H.W., Keener, H.M., Deng, W., Michel, F., 2004. Development and Validation of 3-D

Models to Simulate Airflow and Ammonia Distribution in a High-rise Hog Building

during Summer and Winter Conditions. Agricultural Engineering International CIGR

Journal 6, Manuscript BC 04 044.

Teye, F.K., Hautala, M., 2007. Measuring Ventilation Rates in Dairy Buildings. International

Journal of Ventilation 6 (3), 1473-3315.

Page 90: NEERENGI - Aarhus Universitet

76

Wieringa, J., 1992. Updating the Davenport roughness classification. Journal of wind

engineering and industrial aerodynamics 41-44, 357-358.

Wu, W., Zhang, G., Kai, P., 2012a. Ammonia and Methane Emissions from Two Naturally

Ventilated Dairy Cattle Buildings and the Influence of Climatic Factors on Ammonia

Emissions. Manuscript submitted to Atmospheric Environment.

Wu, W., Kai, P., Zhang, 2012b. An assessment of a partial pit ventilation system to reduce

emission under slatted floor-Part 1: Scale Model Study. Computers and Electronics in

Agriculture 83, 127-133.

Wu, W., Zhang, G., Bjerg, B., Nielsen, P.V., 2012c. An assessment of a partial pit ventilation

system to reduce emission under slatted floor-Part 2: Feasibility of CFD prediction

using RANS turbulence models. Computers and Electronics in Agriculture 83, 134-

142.

Zhang, G., Strom, J. S., Li, B., Rom, H.B., Morsing, S., Dahl, P., Wang, C., 2005. Emission

of Ammonia and other Contaminant Gases from Naturally Ventilated Dairy Cattle

Buildings. Biosystems Engineering 92 (3), 355-364.

Page 91: NEERENGI - Aarhus Universitet

77

Chapter 5

An assessment of a partial pit ventilation system to reduce emission under

slatted floor-Part 1: Scale Model Study

Paper IV:

Wu, W., Kai, P., Zhang, 2012. An assessment of a partial pit ventilation system to reduce

emission under slatted floor-Part 1: Scale Model Study. Computers and Electronics in

Agriculture 83, 127-133.

Page 92: NEERENGI - Aarhus Universitet

78

Abstract

Emissions of ammonia and greenhouse gases from naturally ventilated livestock houses

cause contamination of the surrounding atmospheric environment. Requests to reduce

ammonia emissions from livestock farms are growing in Denmark. It is assumed that using

an additional mechanical pit exhaust unit with a minimised ventilation rate can remove the

most polluted part of the air from the slurry pit and treat it with an air purification unit. This

system can result in reduction of ammonia emissions from naturally ventilated livestock

production units. To study the efficiency of a partial pit ventilation to reduce emissions, a 1:2

scale model of manure pit section of a dairy cattle house was built with slatted floor and a

mechanical pit exhaust at side-walls. Investigations were performed under varied airflow

velocities above the floor, two slatted floor opening ratios (The ratio of the opening area to

the whole floor area), two pit ventilation rates and two exhaust directions. CO2 was used as

tracer gas and was added to the mixing chamber of the pit model with a constant flux. The

removal ratio was defined as the percentage of the gases removed by the pit exhaust. The

results showed that a partial pit exhaust system could abate gas emissions from the slurry pit.

The performance of the system was influenced by the following factors: airflow velocities

above the floor, the slatted floor opening ratio, the pit ventilation rates and the pit exhaust

position. For downwind exhaust, the removal ratios were decreased from about 80% to 50%

when the air velocity above the floor was increased from 0.78 to 1.94 m s-1. The mean of the

removal ratios was 83.1% for all upwind exhaust cases and some downwind exhaust cases.

There was no clear velocity dependency. Lower floor opening ratio could reduce more

emission than the higher opening ratio for most of the cases. Higher pit ventilation rates

resulted in higher removal ratios for most cases of downwind exhaust but did not always give

higher removing efficiency for upwind exhaust. Overall, the upwind exhaust can discharge

8% more CO2 than the downwind exhaust. Removal capability was better correlated to the

four factors (airflow velocities above the floor, the slatted floor opening ratio, the pit

ventilation rates and the pit exhaust position) combined when compared to the correlation to

each individual factor.

The results from the scale model measurements need to be validated by full-scale

experiments.

Key words: Ammonia Pit ventilation; slatted floor; scale model; wind tunnel; livestock; tracer gas

Page 93: NEERENGI - Aarhus Universitet

79

Nomenclature B width of the scale model (m) Rr removal ratio (%)

backC background CO2concentration

TIV the integrated variable

( mg m-3)

exhC CO2 concentration in the pit

HU air velocity at 0.15 m height exhaust duct ( mg m-3) above wind table (m s-1)

iC the concentration at different

pressure difference of upstream and positions ( mg m-3) downstream of the orifice (Pa)

C

concentration difference Subscripts

(mg m-3) back background G CO2 generation rate (mg h-1) exh pit exhaust duct H height of headspace of pit (m) i different positions

PEP pit exhaust position fo floor opening ratio

(-1~upwind, 1~downwind) H height

exhQ pit ventilation rate (m3 h-1)

foR slatted floor opening ratio

Page 94: NEERENGI - Aarhus Universitet

80

5.1. Introduction

Emissions of ammonia and greenhouse gases from naturally ventilated livestock houses cause

the problem of atmospheric pollution. Ammonia is also known to damage ecosystems via

eutrophication (Monteny and Erisman, 1998.). In Denmark the requests for reduction of

ammonia emission from livestock farms is growing (Saha et al, 2010.). In order to reduce the

negative impact of the emission on the environment, the polluted air should be purified

before released into the atmosphere.

In a naturally ventilated livestock building, it is difficult to collect the exhausted air for

further air cleaning. However, partial pit ventilation may be adapted into the system, i.e.,

remove a part of the most polluted air under slatted floor and connect the exhaust channel to

an air purification unit. Saha et al. (2010) investigated the effects of a partial pit ventilation

system on ammonia emission from a fattening pig room with a diffuse ceiling inlet, a ceiling-

roof-top air exhaust and a pit exhaust. The pit exhaust operated with only 10% of the total

ventilation capacity and the ceiling-roof-top exhaust operated as the major ventilation unit

regulated according to the thermal conditions in the room. Saha et al. (2010) concluded that a

partial pit exhaust with purification system could reduce total ammonia emission significantly

even though air exhausted through the pit constituted a relatively small share of the total

ventilation capacity.

Challenges of applying partial pit ventilation system to naturally ventilated dairy cattle

buildings are the wider slatted floor area above the slurry channel and also the more complex

airflow characteristics above the slatted floor due to natural wind effects. It brings up the

issues on how to design a suitable system for such an application.

The hypothesis is that using an additional mechanical pit exhaust unit with a minimised

ventilation rate can remove polluted air and treat them with an air purification unit, which can

reduce ammonia emission from naturally ventilated livestock production systems. The

objectives of this study are:

(1) to validate this hypothesis in a laboratory condition using a scale model and to

investigate the ability of a partial pit ventilation system to reduce gas emissions.

(2) to investigate the effects of the opening ratios of the slatted floor, the pit

ventilation rates and the pit ventilation positions on the capacity of partial pit ventilation

to reduce gas emissions.

5.2. Materials and Methods

Experimental investigations were carried out in Air Physics Lab, Engineering Centre

Bygholm, Aarhus University, Denmark.

5.2.1. Experimental facility

5.2.1.1. Scale model for slurry pit

A one-half scale model of manure pit section was built of wood materials and was composed

Page 95: NEERENGI - Aarhus Universitet

81

of slatted floor and slurry pit beneath (Fig. 5.1-(a)). The size of the scale model was

1.5m×1.25m×0.5m. The height of the head space was 0.4 m and at the bottom of the pit, there

was a tracer gas mixing chamber of 0.1 m in height (Fig. 5.1-(a)). The detailed dimensions of

the scale model are shown in Fig. 5.1-(c). The mixing chamber had a loft plate with 63 evenly

distributed holes (Fig. 5.1-(c)) and covered with porous membrane (Fig. 5.1-(a) (c)). The

plate and membrane were used to release gases. The diameter of each hole was 0.006 m and

the distance between each hole was 0.142 m. Two exhaust openings (Fig. 5.1-(a) (c)) were

placed at the each end of the model. The centres of the openings were in the middle height of

the pit headspace. The diameter of each opening was 0.10m. On the top of the headspace was

a slatted floor (Fig. 5.1-(b)). The dimensions of the slats are shown in Fig. 5.1-(c). The width

of each slot on the floor was 0.02m.

(a) (b)

(c)

Fig. 5.1- The experimental facility. (a) The photograph of the scale model; (b)The photograph of the wind tunnel system; (c)The dimensions of the scale model and the slats. All the units

are in mm.

5.2.1.2. Wind tunnel

The size of the wind tunnel is 11m×2.8m×2.8m. Initial measurement showed a minor

difference in air velocity above two side edges (along the length of the wind tunnel) of a wind

table when air velocity is smaller than 0.4m/s.

Page 96: NEERENGI - Aarhus Universitet

82

A wind table was placed in the wind tunnel to simulate solid floors in naturally

ventilated cattle houses (Fig. 5.1-(b)). The size of the wind table is 6m×2m.The pit model

was placed in the centre of the table and the top of the slatted floor was at the same level with

the top of the table. The slats (Fig. 5.1-(b)) of the pit model were orientated parallel to the

airflow direction in the wind tunnel to simulate a case of cross ventilation in a dairy cattle

buildings.

5.2.1.3. Pit exhaust system

Part of the released gas from the mixing chamber was removed by the pit exhaust system

(Fig. 5.2) and the rest escaped the headspace through the slatted floor slots. Two exhaust

positions were used at the two ends of the model to simulate the exhaust at two sides of slurry

pit channel in a practical dairy cattle building. The normal directions of the exhaust opening

were parallel to the slats on the floor. When the pit exhaust duct was connected to the

openings at the side of upstream of the wind, it was specified as upwind exhaust; otherwise, it

was specified as downwind exhaust (Fig. 5.2).

Fig. 5.2 - The experimental system

5.2.1.4. Tracer gas generation

To simulate emission of ammonia and greenhouse gas from the manure surface in the pit,

carbon dioxide (CO2) was supplied to the mixing chamber via a set of distributed tubes and

the release openings of the tubes were evenly placed in the bottom of the chamber (Fig. 5.2).

A mixing fan was used in the chamber to make sure that CO2 was mixed homogeneously

(Fig. 5.2). The plate with uniformly spaced holes and the covered porous membrane created a

uniform distribution surface for CO2 released from the mixing chamber to the headspace of

the pit.

CO2 was supplied from a pressurized gas cylinder and controlled via an adjustable

valve and a flow meter (Fig. 5.2). A constant flux was kept for all experimental treatments.

Page 97: NEERENGI - Aarhus Universitet

83

5.2.2. Experimental set-up and measurements

5.2.2.1. Experimental set-up

Investigations were performed at five airflow velocities above the floor, at two pit ventilation

rates and at two slatted floor opening ratios (The ratio of the opening area to the whole floor

area) in a wind tunnel. The varied air velocities above the slatted floor were reached by

adjusting the operation frequency of the wind tunnel fan motor. The velocities at 0.15 m

height above the wind table in front of the slatted floor are listed in Table 5.1. The used pit

ventilation rates were 120 and 60 m3 h-1. Two floor opening ratios (21% and 9.7%) and two

pit exhaust positions (downwind and upwind exhaust) were used.

Table 5.1. The velocities at 0.15 m height from the wind table Frequency (Hz) Velocity (m/s) Accuracy (±, m/s) 2 0.19 0.01 5 0.49 0.01 9 0.78 0.01 14 1.46 0.01 20 2.06 0.02

* Frequency is the frequency of wind fan motor

5.2.2.2. Measurement of the velocities and the pit ventilation rates

Air velocity at 0.15 m height above the floor level and at upwind of the pit model was

monitored by Testo 400 (Testo AG, Lenzkirch, Germany) velocity sensor. The monitor

position was at the symmetrical plane (aligned with length of wind tunnel) of the wind tunnel

table. The sensor has a measurement range of 0-20 m s-1 and an accuracy of ±0.01 m s-1 (0-

1.99 m s-1), ±0.02 m s-1 (2-4.9 m s-1) and ±0.04 m s-1 (5-20 m s-1).

Air velocity inside the headspace was measured by an ultrasonic anemometer –

WindMaster (Gill instruments Ltd, Hampshire, UK). It has a velocity range of 0-45 m s-1 and

a resolution of 0.01 m s-1. The direction range is 0-359.9° and the resolution is 1°. The sample

rate can be 20 Hz and 32 Hz. 20 Hz with an output rate of 1 Hz was used in this

measurement. Velocity at each measurement point was recorded for 5 min. The unit is

designed not to require re-calibration within lifetime (Anonymous, 2009). According to

standard calibration, it has accuracy at 12 m s-1 of 1.5% RMS for wind speed and 2° for wind

direction. The head of ultrasonic is 233.5 mm high which is nearly half of the height of

headspace. Thus only velocities at the height of 0.2 m above bottom surface of the headspace

were measured. The ultrasonic sensor can be moved along central slot of the slatted floor to

record velocities at different sites.

A flow meter, type FMU/FMDRU 100-80 (Lindab Ventilation, Båstad, Sweden) was

used to measure pit ventilation rates. The accuracy is 5%.The ventilated flow in the pit

exhaust duct is determined using the equation

PQexh 84.15 (5.1)

Page 98: NEERENGI - Aarhus Universitet

84

where, exhQ is the pit ventilation rate, m3 h-1; 15.84 is a constant from manufactory; P is

pressure difference between upstream and downstream side of the orifice, Pa. The pressure

differences were measured by a pressure sensor (Model 694, Huba Control, Würenlos,

Switzerland) with a range of 0 – 300 Pa, an accuracy of ±0.7% and a resolution of 0.1% of

full dimension. The data was collected through a data logger (CR1000 Campbell Scientific,

Utah, USA). The sampling frequency was 0.1 Hz and the data was recorded as averages of

each one minute.

5.2.2.3. The measurement of CO2 concentration

A constant CO2 flux of 5.76 x 104 mg h-1 was injected into the mixing chamber and monitored

by a flow meter (FM-487 variable area rotameter, Porter Instrument, Hatfield, PA, USA). The

total injection rate was determined by the following method. Pit ventilation was started to

discharge the total generated CO2. The top of the slatted floor was tightly covered but a very

narrow slot was left open at the opposite side of the pit exhaust openings. The narrow slot

was used to introduce fresh air. The total CO2 injection rate can be decided by the following

equation:

)( backexhexh CCQG (5.2)

where, G is CO2 generation rate, mg h-1; Qexh is pit exhaust ventilation rate, m3 h-1; exhC is

CO2 concentration in the pit exhaust duct, mg m-3; backC is background CO2 concentration,

mg m-3.

CO2 concentration was measured with a 1312 Photoacoustic Multi-gas Monitor and a

multiplexer 1303 (Innova air Tech Instruments A/S, Denmark). A reference measurement

point was placed in the centre of the mixing chamber to monitor the gas balance of each

experiment setup. When CO2 concentration at this point became constant, the system was

perceived to be at equilibrium. The background concentration of the airflow above the slatted

floor was measured in the upwind side of the wind tunnel, about 2 m away from the pit side

wall. The concentrations were also measured in the exhaust duct and in three positions X 1,

X 2, X 3 (Fig.1-d) at the half height of the headspace. All the three positions were located in

the symmetric plane of the pit model and parallel to the slots of the floor. After about 2 h

when the system achieved a steady-state condition, the data was sampled for 20 min. The

sampling period for each measurement was 20 s, followed by 20 s cleaning time to replace

the air in the measuring chamber of the gas monitor before a new measurement channel

started to sample.

5.2.3. Estimation of air exchange rate using removal ratio

Removal ratio, defined as the ratio of CO2 flux exhausted by the pit ventilation and the

Page 99: NEERENGI - Aarhus Universitet

85

injected CO2 flux, was used to evaluate the performance of the pit ventilation system under

the different external conditions. It can be expressed as

100)(

G

CCQRr backexhexh (5.3)

where, Rr is removal ratio, %. The larger the Rr, the more CO2 is exhausted through the pit

ventilation, thus the more efficient the pit ventilation is.

5.2.4. The integrate variable to evaluate the effect of all factors on the performance of pit

ventilation

There are four factors: air velocity above the wind table ( HU ), opening ratio of slatted floor

( foR ), pit ventilation rate ( exhQ ) and pit exhaust position, which influenced the removal ratio

of the system. The four factors may have a synergistic effect on the performance of the pit

ventilation and there is no consistent relationship between removal ratio and each individual

factor. Based on this hypothesis, a new index normalized by all the four factors was proposed

as follow and named as the integrated variable ( IV ).

PEPRQBHUIV foexhH )/( (5.4)

where, B is width of the scale model; H is the height of headspace of the scale model; foR is

slatted floor opening ratio; PEP was used to quantify the pit exhaust position. It could be

described as

nventilatiopitdownwind

nventilatiopitupwindPEP

1

1 (5.5)

5.2.5. CO2 concentration difference

CO2 concentration difference between pit head space and background was calculated from

backgroundi CCC (5.6)

Where C is concentration difference, mg m-3 ; iC is the concentration at different positions

X1, X2 or X3 (Fig.2), mg m-3; i represents position X1, X2 or X3.

5.3. Results

5.3.1. CO2 concentration inside the pit headspace

C and the standard error are given in Table 5.2. The averageC at the three positions was

2419±371 mg m-3 for the case without exhaust and this concentration difference was reduced

to 1782±278 and 1567±285 mg m-3 when the downwind pit ventilation rate was 60 and 120

m3 h-1, respectively. In the cases of upwind exhaust, C was largely reduced to 243±78 and

62±19 mg m-3 for pit ventilation rate 60 and 120 m3 h-1.

Page 100: NEERENGI - Aarhus Universitet

86

Table 5.2. CO2 concentration difference between pit headspace and background (mg m-3) Position Without exhaust Downwind exhaust Upwind exhaust

Ventilation rate (m3 h-1) Ventilation rate (m3 h-1)

60 120 60 120

Mean SEa Mean SE Mean SE Mean SE Mean SE

X1b 1034.40 27.79 665.06 28.50 585.78 27.59 60.02 4.20 10.46 1.12

X2 2249.68 392.68 1745.56 446.23 1041.22 60.84 103.80 10.80 10.20 4.34

X3 3973.24 447.30 2937.80 163.90 3075.30 183.92 567.18 158.46 166.84 2.79

Meanc 2419.11 371.02 1782.81 278.57 1567.43 285.48 243.66 78.52 62.50 19.79

CO2 concentration was measured when the air velocity above the floor was 0.78 m s-1 and the slatted floor opening was 21%.

aSE standard error bsee Fig.1- (d) for different positions

cAverages of the concentration difference at three positions

It should be pointed out thatC and the related standard errors were higher in the upwind

side (X3). In the case without exhaust, C was 1034±62 mg m-3 in the downwind side (X1)

and it was 3.84 times higher in the upwind side (X3).

5.3.2. The effect of the air velocity

Fig. 5.3 shows that the removal ratios were affected by the pit ventilation rates and positions,

slatted floor opening ratio and the air velocities above the floor. When UH was increased from

0.78 to 1.94 m s-1 for the cases of downwind exhaust, the removal ratios decreased from about

80% to 50%. The mean of the removal ratios was 83.1% for all the cases of upwind exhaust

and some cases of downwind exhaust when UH was smaller than 0.78 m s-1. The removal

ratios varied from about 70% to 98% but there was no clear velocity dependency.

(a) (b)

Fig. 5.3 – The change of removal ratio with airflow velocities at 0.15 m above the wind table: (a) downwind exhaust and (b) upwind exhaust. 120 and 60 represent the two ventilation rates

in m3 h-1; and 21% and 9.7% represent the two floor opening ratios.

5.3.3. The effect of the slatted floor

In all cases of downwind exhaust, the lower floor opening ratio (9.7%) resulted in of 20% or

Page 101: NEERENGI - Aarhus Universitet

87

more gas removing in average than the higher opening ratio (21.0%) (Fig. 5.3-a).

For the upwind exhaust, the lower floor opening ratio (9.7%) resulted in of 10% or more gas

removing than the higher opening ratio (21%) for most cases of exhQ =60 m3 h-1. When the pit

ventilation rate was 120 m3 h-1, the lower opening ratio (9.7%) did not result in higher

removal ratios (Fig.3 - b); in this situation, lower and higher opening ratio exhausted 77.9%

and 78.5% emission averagely by pit ventilation system, respectively.

5.3.4. The effect of the pit ventilation configurations (Ventilation rate and position)

Higher ventilation rates caused higher removal ratios for most cases of downwind exhaust

(Fig. 5.3a). Generally, using the higher pit ventilation rates (120 m3 h-1) removed on average

about 79% CO2 and using the lower ventilation rates (60 m3 h-1) removed on average about

67% CO2.

For the upwind pit exhaust, there was not such a trend as in downwind side. When floor

opening ratio was 21%, higher ventilation rates (120 m3 h-1) did not remove more emission,

compared with lower ventilation rates (120 m3 h-1). Both ventilation rates exhausted about

78% CO2. When floor opening ratio was 9.7%, lower ventilation rates (60 m3 h-1) removed on

average79% CO2; higher ventilation rates (120 m3 h-1) removed on average 89% CO2.

Pit ventilation position had a significant influence on the removal capability (Fig. 5.3).

Overall, upwind exhaust discharged 8% more CO2 through the pit ventilation system. When

the air velocity was lower than 0.49 m s-1, removal ratios did not change much with exhaust

positions. When the air velocity was higher than 0.49 m s-1, about 15% more CO2 could be

exhausted by changing downwind to upwind pit ventilation. The minimum removal ratios

were 31.7% and 66.7% for downwind and upwind exhaust, respectively. Ventilation position

on upwind side can abate emission more significantly, when compared to that on the

downwind side.

5.3.5. The overall effect of the four factors

Fig. 5.4 shows the relationship between the removal ratio and the integrated variable IV.

When IV was higher than 1.91, the higher IV resulted in lower removal ratios (Fig. 5.4). The

removal ratio decreased with the increase of the velocity, with the decrease of the pit

ventilation rate and with the decrease of the slatted floor opening areas. When IV was lower

than 1.91 and higher than -12, the removal ratios did not change according to the variation of

IV.

5.4. Discussion

5.4.1. CO2 concentration inside the pit headspace

Higher CO2 concentration inside the pit headspace could lead to higher emission into the

room space due to convection and diffusion. The highest CO2 concentration under the slatted

floor occurred in the case without pit ventilation system and the concentration was

Page 102: NEERENGI - Aarhus Universitet

88

significantly reduced with the application of pit exhaust. The results showed the ability of the

partial pit ventilation to reduce the gases under the slatted floor.

Fig. 5.4 – Removal ratios from measurement under the influence of different pit ventilation

rates, airflow velocities above the wind table and two slatted floor opening ratios.

The concentration on upwind side was higher than that on downwind side (Table 2) by

a factor of 2~3 when the air was exhausted on the downwind side, comparable to the findings

of Liu and Barth (2001). The reason might be that the airflow above the slatted floor entered

the pit mainly from the downwind side; part of it was exhausted directly by pit ventilation

system; part of it attached to the downwind side wall and re-circulated to the upwind side.

CO2 was thus accumulated in the upwind side and caused high concentration. Johnson and

Hunter (1998) reported a similar airflow type in a study of flow over street canyon, which

was similar to the flow over a pit.

The standard errors of concentration were very large, compared with corresponding

mean values. This showed high fluctuations of CO2 concentration. The unsteady behavior of

pollutant distribution was also reported in the work of Sagrado et al, (1998).

5.4.2. The effect of air velocity

The air velocity above the floor affected the air exchange rates in the headspace of the slurry

pit. Higher air velocity (2 m s-1) can mix the entrained fresh air and the injected CO2 better.

Consequently, more pollutants could be brought out by advection and diffusion.

Liu and Barth (2001) found that pollutant in a cavity was transported to free stream

mainly by turbulence. When the air velocity above the floor was increased, the Reynolds

number in the floor opening was increased; consequently, the airflow inside pit became more

turbulent. More CO2 was transported into the wind tunnel space from upwind side. This could

explain that the removal ratios were decreased by the increase of the air velocity ( HU >0.78

m s-1) above the floor for downwind exhaust. When air velocity was lower than 0.78 m s-1,

the Reynolds number in the floor opening was very small and the flow regime could be

between laminar and turbulence. The transient flow could cause the instability of the airflow

Page 103: NEERENGI - Aarhus Universitet

89

patterns. This could explain that the removal ratios varied around 83.1% when air velocity

was smaller than 0.78 m s-1for both downwind and upwind exhaust.

5.4.3. The effect of the slatted floor

Fig. 5.3 shows that smaller slatted floor opening ratio resulted in lower emissions in most of

the cases. The results were similar to those reported by Zhang et al. (2008). It was

straightforward that smaller floor opening area could confine more CO2 under the slats than

larger floor opening area. In some upwind exhaust cases ( exhQ =120 m3 h-1 and

HU =0.78~1.49 m s-1), the mean values of removal ratios with lower floor opening ratio

(9.7%) were smaller than that with higher floor opening ratio (21%); however, the removal

ratios for the two floor openings were not significantly different (p>0.15) during this

condition.

5.4.4. The effect of pit exhaust configurations

The fact that CO2 concentrations in the pit headspace using the upwind exhaust were much

lower than that using the downwind exhaust (Table 5.2) proved that the most polluted air in

the vicinity of the upwind side was discharged directly to outside by using the upwind

exhaust. Upwind exhaust avoided the recirculation of CO2 inside the pit thus reducing the

chance of being emitted to the free stream. This means that upwind exhaust can remove more

emission from the pit headspace than downwind exhaust.

As expected, higher pit ventilation rates removed more CO2 from the pit. However, in

some cases of upwind exhaust, more gas was removed by lower pit ventilation rates (60 m3 h-

1). It could be explained as follow. For upwind exhaust, pressure of upwind side of the pit

became negative. The negative pressure produced a drag force. As a result, the part of the

airflow above the floor did not have enough energy to travel to downwind side wall of the pit.

It entrained the headspace from somewhere near upwind side and was discharged through

exhaust system. Therefore, a vortex was formed close to upwind side. Meanwhile, a second

vortex was formed in the downwind side. A similar airflow pattern containing two vortices in

a cavity can be found in the work of Chang and Meroney (2003). Under this airflow

distribution, the presence of the second vortex would give pollutant more chance to be

transported to the free stream. When the pit ventilation rate was increased, the negative

pressure difference was increased and thus the position of the airflow entering the pit would

be closer to the upwind side wall of the pit. The second vortex would become larger. More

pollutant would be involved in the second vortex; consequently, less pollutant would be

removed by the pit exhaust. Therefore, increased pit ventilation rates could probably not

remove more contaminant in such a case.

Page 104: NEERENGI - Aarhus Universitet

90

5.4.5. The overall effect of all the factors

Fig. 5.4 shows that the removal ratios in downwind exhaust conditions (IV≦1.90) and in all

upwind exhaust conditions varied around the mean value of 83.1% (with a standard deviation

8.8%). According to statistical analysis, the removal ratios were normally distributed.

Therefore, %1.83Rr could be used to predict the removal ratios in the previous conditions.

In most of the cases, the removal ratios were close to each other when the values of

corresponding IV were close. It can be concluded that the removal ratios were more linked to

IV rather than each individual factor.

The concept of IV can help engineers to design partial pit ventilation system. Fig.4

shows that when the air velocity above the floor was small ( HU <0.78 m s-1) or the directions

of the pit exhaust and airflow above the floor were opposite, the overall removal capacity of

the system was nearly constant. The worst removal capacity occurred when the direction of

the airflow was the same with that of pit exhaust. The determination of the pit ventilation rate

could be based on the possibly worst removal capability.

In a naturally ventilated livestock house, the airflow condition near the slatted floor will

be much more complicated than that in the wind tunnel. Airflow patterns above and under the

slatted floor were also affected by the thermal conditions (Zhang et al., 1996.). Furthermore,

differences in scale may affect emissions because of different turbulence scales and

concentration boundary layers near the floor (Saha et al., 2011.). Therefore, the performance

of the partial pit ventilation to reduce emission needs to be investigated and validated by full-

scale experiments.

5.5. Conclusions

Partial pit ventilation is able to remove a large portion of polluted gases under the slatted

floor.

When the air velocity was increased from 0.78 to 1.94 m s-1 for the cases of downwind

exhaust, the removal ratios were decreased from about 80% to 50%. The mean of the removal

ratios was 83.1% for all the cases of upwind exhaust and some cases of downwind exhaust

and there was no clear velocity dependency.

In most cases, the lower floor opening ratio (9.7%) could reduce more gas than the

higher opening ratio (21.0%).

Higher ventilation rates caused higher removal ratios for most cases of downwind

exhaust.

For the upwind side, higher ventilation rates (120 m3 h-1) did not always remove more

emission.

Pit ventilation position had a significant influence on the removal capability. Overall,

upwind exhaust can discharge 8% more CO2 by pit ventilation system than downwind

exhaust.

Page 105: NEERENGI - Aarhus Universitet

91

Removal ratios were more linked to the integrated effect of all the factors rather than

each individual one.

The work could help to design pit ventilation system and decide rational ventilation

rates. The results from the scale model measurements need to be validated by full-scale

experiments.

References

Anonymous, 2009. User Manual: WindMaster & WindMaster Pro Ultrasonic Anemometer.

Doc No. 1561-PS Issue 04. Gill Instruments Limited, Saltmarsh Park, 67 Gosport

Street, Lymington, Hampshire SO41 9EG, UK.

Chang, C.H., Meroney, R.N., 2003. Concentration and flow distributions in urban street

canyons: wind tunnel and computational data. Journal of wind engineering and

industrial aerodynamics 91, 1141-1154.

Monteny, G. J., Erisman, J. W., 1998. Ammonia Emission from Dairy Cow Buildings: A

review of measurement techniques, influencing factors and possibilities for reduction.

Netherlands Journal of Agricultural Science 46, 225-247.

Johnson, G.T., Hunter, L.J., 1998. Urban wind flows: wind tunnel and numerical simulations

—a preliminary comparison. Environmental Modelling and Software 13, 279–286.

Liu, C.H., Barth, M.C., 2001. Large-Eddy Simulation of Flow and Scalar Transport in a

Modeled Street Canyon. Journal of Applied Meteorology 41, 660-673.

Sagrado, A.P.G., Beeck, J.V., Rambaud, P., Olivari D., 2002. Numerical and experimental

modelling of pollutant dispersion in a street canyon. Journal of Wind Engineering and

Industrial Aerodynamics 90, 321-339.

Saha, C.K., Zhang, G., Kai, P., Bjerg B., 2010. Effects of a partial pit ventilation system on

indoor air quality an ammonia emission from a fattening pig room. Biosystems

Engineering 105 (3), 279-287.

Saha, C.K., Wu, W., Zhang, G., Bjerg B., 2011.Assessing effect of wind tunnel sizes on air

velocity and concentration boundary layers and on ammonia emission estimation using

computational fluid dynamics. Computers and Electronics in Agriculture 78, 49-60.

Zhang, G., Morsing, S., Strom, J. S., 1996.Modeling jet drop distances for control of a non-

isothermal, flap adjusted ventilation jet. Transactions of the ASAE 39 (4), 1421-1431.

Zhang, G., Strom, J. S., Li, B., Rom, H.B., Morsing, S., Dahl, P., Wang, C., 2005. Emission

of Ammonia and other Contaminant Gases from Naturally Ventilated Dairy Cattle

Buildings. Biosystems Engineering 92 (3), 355-364.

Zhang, G., Bjerg, B., Strøm, J.S., Morsing, S., Kai, P., Tong, G., Ravn, P., 2008. Emission

effects of three different ventilation control strategies - a scale model study. Biosystems

Engineering 100 (1), 96-104.

Page 106: NEERENGI - Aarhus Universitet

92

Chapter 6

An assessment of a partial pit ventilation system to reduce emission under slatted floor-Part 2: Feasibility of CFD prediction using RANS turbulence models

Paper V:

Wu, W., Zhang, G., Bjerg, B., Nielsen, P.V., 2012. An assessment of a partial pit

ventilation system to reduce emission under slatted floor-Part 2: Feasibility of CFD

prediction using RANS turbulence models. Computers and Electronics in Agriculture 83,

134-142.

Page 107: NEERENGI - Aarhus Universitet

93

Abstract

CFD simulations were carried out to assess the feasibility of using RANS (Reynolds -

averaged Navier – Stokes) turbulence models to evaluate the performance of a partial pit

ventilation system to reduce gas emission under slatted floor. The entire system included a pit

model with slatted floor and pit exhaust system, a wind table to simulate ground and a wind

tunnel to simulate room space of a naturally ventilated livestock house. CFD simulations

started with the selection of a proper domain. Two domains were chosen to evaluate the

effect of domain simplification. The results showed that the effect was significant. The

assessment of different turbulence models including the standard, RNG, realizable

k models; transition SST k model; Reynolds Stress Models were conducted. Results

of RSM were found out to agree best with measured results. In order to understand the

transportation mechanism of pollutant through slatted floor, vertical mean and turbulent flux

were defined and calculated. It was found that turbulence diffusion dominates the

transportation of pollutant from the pit headspace into the free stream. Although

discrepancies existed for some conditions, good agreements of measured and calculated

removal ratios were found for most of cases. It was feasible to use RSM to predict the

removal capability.

The future research should focus on depicting the airflow patterns inside the pit by

using scale model under well controlled laboratory conditions and generating benchmark

data to validate and improve CFD methods. Unsteady CFD simulation using large eddy

simulation could be conducted to make more precise predictions of removal capability of the

partial pit ventilation system by considering the unsteady phenomena of gas emission.

Key words: CFD; RANS; pit ventilation; slatted floor; turbulence diffusion

Page 108: NEERENGI - Aarhus Universitet

94

Nomenclature

b power of the power-law U the velocity (m s-1) function, 0.22 W average vertical velocity (m s-1)

C average concentration of CO2,

X X coordinate, aligned with the

(g s-1 m-3) length of the wind tunnel (m) 'c

the deviation from the mean Y

Y coordinate, aligned with the concentration ( mg m-3) width of the wind tunnel (m)

C constant, 0.09 Z Z coordinate, from wind tunnel

mF vertical flux of CO2 by mean ground to its top wall (m)

flow (g m-2 s-1) HZ the reference height, 0.15 m

tF vertical flux of CO2 by z the height from wind table, m turbulent flow (g m-2 s-1)

'w the deviation from the mean

cK coefficient of scalar transport vertical velocity (m s-1)

exhQ pit ventilation rate (m3 h-1) k turbulent kinetic energy

foR slatted floor opening ratio the dissipation rate Rr removal ratio (%) Subscripts

tSc turbulent Schimidt number, 0.9 fo floor opening ratio

TIV the integrated variable H height

HU air velocity at 0.15 m height

above the slatted floor (m s-1)

Page 109: NEERENGI - Aarhus Universitet

95

6.1. Introduction

A one-half scale model of manure pit section was built and placed in a wind tunnel to

simulate the emission conditions from the slurry pit of a naturally ventilated dairy cattle

house. The scale model contained a headspace and a mixing chamber, which were separated

by a plate with uniformly spaced holes and a porous membrane. CO2 was supplied to the

mixing chamber beneath the pit headspace. The scale model was equipped with a slatted floor

and a mechanical pit exhaust system. The objective of this system is to study the ability of a

partial pit ventilation system to reduce gas emission under slatted floor. A complete

description of the experimental set-up and results can be found in the work of Wu et al.

(2012).

Although the pit model can be placed in a wind tunnel to investigate the performances

of a pit exhaust under controlled laboratory conditions, it is still difficult to observe and

visualise the airflow patterns in the pit headspace. Therefore, the mechanism of gas

transportation through the slatted floor cannot be well understood. Additionally, the influence

of the air velocity above floor, the floor opening ratios and pit ventilation configurations on

the performance of the pit ventilation system was investigated under limited scenarios. It was

due to the fact that the wind tunnel measurements were relatively expensive and time

consuming. The possibility of using Computational Fluid Dynamics (CFD) was studied in

this work and CFD was intended to supplement the limitations of scale model.

CFD has become a useful and popular tool to predict airflow characteristics and assess

gas emissions across wide research areas: airflow distribution in greenhouses (Bartzanas et al,

2002); ammonia emission from pig houses (Rong et al, 2010); dynamic flux chamber

methodology (Saha et al, 2011). CFD simulation has also been performed to evaluate the

efficiency of a partial pit ventilation system to reduce ammonia emission in pig units with

animals and slatted floor ( Bjerg et al, 2008a) and in a naturally ventilated cattle building (

Bjerg and Andersen, 2010).

In the work of Bjerg et al (2008a), the slatted floor was handled in a form of porous

media. However, the effect of the slatted floor on the mechanism of emission transportation

should be critical and the slatted floor should be modelled in geometrical details. For the

application of pit ventilation in a naturally ventilated building ( Bjerg and Andersen, 2010), a

two dimensional model was adopted and the results were not validated with measurements.

Research related to the application of CFD on partial ventilation system is still necessary in

order to consider the three dimensional effects, validation with measurements and the

detailed simulation of the slatted floor. No report on this kind of simulation has been found in

literature.

So far, most CFD models have been successfully used for airflow around bluff

structures (Castro and Apsley, 1997) and airflow inside confined spaces (Zhai, 2007). But it

Page 110: NEERENGI - Aarhus Universitet

96

is not sure if the current turbulence models and numerical methods can be applied to the case

that involves airflow through cavities (the pit headspace) and interacting with the free stream

through narrow slots. Studies on airflow over cavities can be found in the field of urban wind

flows (Li et al, 2006), where research was focused on the pollutant dispersion in urban street

canyons.

Different turbulence models were evaluated to study urban wind flows. A two-

dimensional numerical investigation by Huang et al. (2000) showed a good agreement

between the calculated and observed concentrations, which made it possible to use the

standard k turbulence model (Launder and Spalding, 1974) to analyze the pollutant

distributions emitted from vehicles within urban street canyons. Sagrado et al., 2002 showed

that the realizable k model (Shih et al., 1995) with a two-layer zonal approach to a wall

had been found to be the most accurate one in k models; the conclusion was achieved by

comparing the results with direct numerical simulation on separated flows and flows with

complex secondary flow features in street canyons. A three-dimensional simulation with the

RNG k turbulence scheme (Choudhury, 1993) by Kim et al.(2004) indicated that the main

features of mean flow were simulated well although the velocities were underestimated in

some parts of the street canyons. A work presented by Sahm et al. (2002) pointed out that the

velocity discrepancies among results of different k turbulence models and measurements

were not negligible at individual locations, even though the case was relatively simple.

According to the review of the research on the CFD simulation of urban wind flow,

different turbulence models were found to be suitable for different cases and discrepancy

between simulation and experiment was also found to be negligible. The flow in this study

should be more complex because of the involvement of pit exhaust system and the slatted

floor. The slatted floor makes the boundary layer above and under the floor rather difficult to

model. Turbulence models will be crucial for an accurate simulation of the flow in slots and

headspace.

The objective of this study is to assess the feasibility of using RANS (Reynolds -

averaged Navier - Stokes) turbulence models to evaluate the performance of a partial pit

ventilation system to reduce gas emission under slatted floor. The accuracy of the three

k turbulence models (standard, RNG, realizable) (Davidson, 1997), the transition SST

k turbulence model (Menter, 1994) and RSM (Reynolds Stress Models) (Hanjalic, 1999)

were tested to predict the removal ratio, which was defined as the percentage of contaminant

removed by pit exhaust.

6.2. Materials and Methods

A 1:2 scale model of slurry pit section with slatted floor and CFD simulation were used as

tools to study the efficiency of a partial pit ventilation system to reduce emission under

Page 111: NEERENGI - Aarhus Universitet

97

slatted floor. A detailed description of scale model study can be found in the work of Wu et

al. (2012). Set-up of CFD is given in the following sections.

6.2.1. Basic concept of CFD and software

The basic concept of CFD is to solve a set of partial differential equations. The governing

equations are the Navier-Stokes and continuity equations (Launder and Spalding, 1974). CO2

was used as tracer gas in previous experiments (Wu et al, 2010). It was treated as scalar in

this CFD simulation. The transport equation for CO2 is written as (Baik and Kim, 1998):

cccc Sz

CK

zy

CK

yx

CK

xz

CW

y

CV

x

CU

t

C

)()()( (6.1)

where, C is the mean concentration of a passive pollutant, t is the time,U , V and W are the

mean velocities in the X , Y and Z coordinate direction, respectively. cK is the turbulent

diffusivity for the scalar variable and cS denotes the source or sink term of the pollutants.

tc Sc

kCK

2

(6.2)

where, C is a constant and equal to 0.09; k is the turbulent kinetic energy; is the

dissipation rate; Sct is the turbulent Schmidt number and specified as 0.9 (Sini et al., 1996).

Commercial software Fluent 12.0 was used to solve the equations on the basis of finite

volume method. Standard k , k RNG, k Realizable, transition SST k and RSM

turbulence models were tested in this work. All the velocity and turbulence terms for

convection were approximated using second order upwind scheme. Diffusion terms were

discretised using central difference scheme. SIMPLE method was employed for the pressure-

velocity correction.

6.2.2. Computational domain

CFD simulation starts from the specification of the computational domain. The experimental

facility is usually geometrically simplified in order to make the calculation feasible and easy.

To assure the quality of the simulated results, the computational domain should be as close to

the original geometry as possible. Two computational domains were included in this work.

The whole wind tunnel containing the wind table and the scale model was built as the

first computational domain, defined as domain 1 (Fig. 6.1a). The set-up of the first domain

was exactly the same as that in the experiment. The difficulty of using this domain was to

define the boundary condition of the wind tunnel fan. Data of the fan was not available.

Since the measured velocity profiles above the wind table were available, an alternative

was to cut off the spaces under wind table and extend the wind table to the whole wind

tunnel. In this way, only the space above the wind table and the headspace of the scale model

was treated as a domain, defined as domain 2 (Fig. 6.1b). The length, width and height of

Page 112: NEERENGI - Aarhus Universitet

98

domain 2 were the same with domain 1. The measured velocity profiles can be easily used as

the inlet boundary condition.

For both domains, X, Y, Z coordinate was aligned with the length, width and height of

the wind tunnel, respectively.

The simulation results of using the two domains were evaluated in this work.

(a) (b)

Fig. 6.1 – Schematic of the two domains: (a), domain 1; (b), domain 2.

6.2.3. Boundary conditions

For domain 1, the upstream end of the wind tunnel was defined as velocity inlet with fixed

velocity magnitudes. Since only the air velocities above the wind table were known, the

corresponding velocities for the fan were determined by the following procedure. An initial

speed was assigned to the fan for a preliminary calculation. A velocity profile above the wind

table was obtained from the calculation. The initial speed was calibrated by the ratio of the

calculated velocities to those from the measurements. By iterations of calculation and

calibration, a proper air velocity generated by the fan was finally decided. The final velocity

profiles from simulations and measurements are given in Fig. 6.2.

Fig. 6.2 – Velocity profiles of measurements (open circles) and simulations (solid lines) in

the upwind side of the wind tunnel and above the wind table. The simulated velocity profiles

were obtained by using domain 1.

For domain 2, the inflow surface was defined as velocity inlet and assumed to follow a

power-law profile (Tsai and Chen, 2004 ). The power-law function was given by

Page 113: NEERENGI - Aarhus Universitet

99

bHH ZzUU )/( (6.3)

where, U is the velocity, m s-1; z is the height from the wind table, m. b is 0.22 according to

the linear regression of the measured velocities; UH (m s-1) is the air velocity at 0.15 m height

above the wind table and can be found in Table 1 of Wu et al. (2011); ZH is the reference

height, 0.15 m.

The outlet of the wind tunnel was defined as pressure outlet for both domains. The

measured pit ventilation rates were converted to velocities that were used as the input values

of the boundary condition of the exhaust openings. The bottom surface of the headspace of

the pit was specified as wall and appointed as emission surface. The tracer gas was treated as

a scalar quantity in the simulation and 1 kg s-1 m-2 was set as the scalar generation rate of the

emission surface. No scalar distribution was assumed to the whole domain as initial

condition. Other surfaces of both domains were considered as walls.

6.2.4. Mesh

More grids were allocated around the slatted floor and inside the slatted slot. The width of the

slot was 0.02 m. Thus the smallest grid distance was required to be smaller than 0.02 m. The

mesh size of other edges was increased proportionally based on that of the shortest edge.

Hence the coarse grid number was relatively large, which was 1308854 and 739020 for

domain 1 and domain 2, respectively. The medium and fine grid number were 1888437,

2726686 for domain 1 and 1042018, 1490086 for domain 2. The optimum grid distribution

for each domain was achieved by comparing the results of three meshes. Velocity profiles in

the pit headspace from three different meshes for the domain are compared in Fig. 6.3. The

velocities calculated from medium mesh were very close to that from the fine mesh. It proved

that medium mesh was good enough for achieving mesh convergence. In the same manner,

medium mesh for domain 2 was chosen.

Fig. 6.3 – Investigation of mesh convergence: velocity profiles in the pit headspace (from Z

=0.10 m to Z =0.50 m) in the middle of the pit model

Page 114: NEERENGI - Aarhus Universitet

100

6.2.5. Calculation of vertical mean and turbulence flux

An instantaneous velocityu can be decomposed into a mean flow componentU and with a

fluctuation value 'u superimposed on it. Considering a steady case without source, gas is

transported by mean flow (U ) and its fluctuation ( 'u ). The part of gas transported by mean

flow is called mean flux ( mF ); whereas the rest of gas transported by its fluctuation is named

as turbulent flux ( tF ). The investigation of the relationship between mean flux and turbulent

flux would benefit finding the transportation mechanism.

The vertical flux of the scalar by mean and turbulent flow was calculated using (Baik

and Kim, 2002)

CWFm (6.4)

z

CKwcF ct

'' (6.5)

where, C is the mean concentration of the scalar, W is the vertical mean velocity; 'c is the

deviation from the mean concentration and 'w is the deviation from the mean vertical

velocity. cK is the turbulent diffusivity defined in equation (6.2).

6.3. Results

6.3.1. The effect of simplification of computational domains

Domain 2 was aimed to simplify the geometry and to use measured velocity profiles as inlet

boundary condition. Removal ratios for all the cases were higher than 90% by using domain

2. The large discrepancy between simulation and measurement made domian2 unacceptable.

Removal ratios using domain 1 were comparable with those from measurement. Therefore,

domain 1 was selected to conduct all the following simulations.

6.3.2. The effect of different RANS turbulence models

Table 6.1 shows the air velocities in the pit headspace and the removal ratios calculated using

different RANS turbulence models and those obtained from measurements. The simulations

were conducted under the same condition ( HU =0.78 m s-1, foR =21.0%, exhQ =120 m3 h-1 and

downwind exhaust) for all the turbulence models. X =4.550 to 5.450 m was upwind side to

downwind side of the pit headspace. The velocities predicted by CFD agreed with the

measurement values very well at upwind side and discrepancy became larger and larger when

the position was approaching downwind side. Generally, the velocities from RSM agreed

better with the measured velocities. An underestimation of Rr was noticed for all the

turbulence models. The transition SST k turbulence model underestimated Rr more than

Page 115: NEERENGI - Aarhus Universitet

101

8.1%. The three k models predicted similar removal ratios (around 75%). The deviations

from the measured values were larger than 5%. The best agreement between CFD simulation

and measurement was achieved by RSM turbulence model. The deviation was less than 2%.

Therefore, RSM was chosen for further calculations of Rr under different conditions.

Table 6.1 Comparison of measured velocity magnitudes and removal ratios with that calculated from different turbulence models

Turbulence models and measurement

Velocity (m s-1) Rr At position (m): Y=1.426, Z=0.3

X=5.450 X=5.150 X=5.000 X=4.850 X=4.550 k Standard 0.133 0.097 0.082 0.065 0.028 76.8% k RNG 0.115 0.09 0.072 0.062 0.036 75.9%

k Realizable 0.128 0.09 0.077 0.07 0.039 74.2% SST k 0.158 0.073 0.073 0.05 0.04 73.1%

RSM 0.112 0.103 0.074 0.062 0.044 79.7%

Experiment 0.103 0.062 0.051 0.058 0.047

81.8% (±0.027) (±0.029) (±0.027) (±0.026) (±0.021)

* Simulations were carried out in the condition of HU =0.78 m s-1, foR =21.0%, exhQ =120 m3

h-1 and downwind exhaust.

6.3.3. Distribution of velocities and concentrations

Two planes in the domain were used to demonstrate the velocity, concentration fields and

airflow patterns. Plane 1 (Fig. 6.4) was aligned with Y coordinate and perpendicular with the

direction of the slats. Plane 2 (Fig. 6.4) was aligned with X coordinate and parallel to the

direction of the slats.

Fig. 6.4 – Schematic of the position Plane 1 and Plane 2

Examples of velocity and concentration distribution are shown in Fig. 6.5 for HU =0.78

m s-1, foR =21.0%, exhQ =120 m3 h-1 at the downwind exhaust. For both planes, the velocity

magnitude was about 0.80 m s-1 in the space above the wind table; the scalar concentration

was below 0.01 kg m-3. Little scalar was transported to the space above the floor. The

velocities on Plane 1 were below 0.1 m s-1 for most parts of the pit headspace (Fig. 6.5a).

Both velocity and concentration fields showed symmetry along the Y middle plane aligned

with the width of the wind tunnel (Fig. 6.5a, c). It can be seen from Plane 2 that the fresh air

Page 116: NEERENGI - Aarhus Universitet

102

travelled from the upwind side edge of the pit to the downwind side edge of the pit; part of

that was exhausted by the pit ventilation system; the rest of the air circulated to the upwind

direction and its return distance was 1/3 of the length of the pit (Fig. 6.5b). The concentration

was high on upwind side of the pit headspace (Fig. 6.5d).

(a) (b)

(c) (d)

Fig. 6.5 – Velocity field of: (a) Plane 1; (b) Plane 2 and Scalar distribution of: (c) Plane 1; (d) Plane 2 in the case of HU =0.78 m s-1, foR =21.0%, exhQ =120 m3 h-1 and downwind exhaust

6.3.4. Airflow patterns

Fig. 6.6 shows that airflow patterns on Plane 1 with varied velocity ( HU ) for

foR =21.0%, exhQ =120 m3 h-1 at the downwind exhaust. When HU was 0.19 m s-1, fresh air

entered the pit headspace through all the slots (Fig. 6.6a). When HU was increased to 2.06 m

s-1, air entrained the middle slots but some air exit to the space above the floor from the slots

near the two edges (parallel to the slot opening) of the pit model (Fig. 6.6b). General

simulation results in all cases showed that airflow patterns on Plane 1 were affected by the air

velocity above the slatted floor UH. No similar effect on airflow pattern in plane 1 was found

by changing foR , exhQ and the exhaust position.

Fig. 6.7 shows that the airflow patterns on Plane 2 varied with exhaust positions

for foR =21.0%, exhQ =120 m3 h-1 and HU = 2.06 m s-1. Both conditions show a similar airflow

pattern: air travelled from the upwind edge to the downwind edge of the pit wall, and

attached to the downwind side wall; part of the flow returned to the upwind direction along

the bottom surface of the pit. The significant difference was the travel distance of the return

Page 117: NEERENGI - Aarhus Universitet

103

flow. Return flow can travel more close to upwind side by using upwind exhaust (Fig. 6.7b).

The overall simulation results in all cases showed that airflow patterns in Plane 2 were

mainly affected by exhaust positions rather than foR , exhQ and HU .

(a) (b)

Fig. 6.6 – Airflow patterns in Plane 1: (a) HU =0.19 m s-1; (b) HU = 2.06 m s-1

when foR =21.0%, exhQ =120 m3 h-1 and the pit ventilation position is at downwind side

(a) (b) Fig. 6.7 – Airflow patterns in plane 2: (a) downwind exhaust; (b) upwind exhaust when

HU = 2.06 m s-1, foR =21.0% and exhQ =120 m3 h-1

6.3.5. Removal ratios predicted by CFD

Removal ratios based on CFD simulations under different cases are presented in Fig. 6.8.

Fig. 6.8 – Removal ratios under the influence of different pit ventilation rates, airflow

velocities at the slatted floor level and two-slatted floor opening ratios. Open circle-

experimental values; open square-simulated values; dash line-fourth order polynomial fitted

values using measured data ( 2R =0.805); solid line-fourth order polynomial fitted values

using simulated data ( 2R =0.895)

Page 118: NEERENGI - Aarhus Universitet

104

In order to display the overall trend of Rr varying with IV, defined by Wu et al. (2012),

the measured and simulated Rr values were fitted using 4th-order polynomial function. The

predicted Rr follow the trend of measured Rr in most of the cases. A large discrepancy was

found when IV was about -11.6, and the Rr predicted by CFD was 20% lower than the

measured values. The discrepancy made the trends of measurement and simulation different

for IV <-7. The simulation results showed a decrease of Rr with the increase of absolute

values of IV. But the decrease for downwind exhaust was fast than that for upwind exhaust.

Good agreement was found between measured and predicted Rr values. When

HU =0.49 and 2.06 m s-1, the discrepancy between simulated and measured removal ratios

was about 10%~20%.

6.3.6. The vertical CO2 flux through the slatted floor openings

Fig. 6.9 shows the horizontal distributions of the calculated vertical mean and turbulent flux

of CO2 through the top surface of the centre slot (see Fig. 6.4). The calculation was based on

HU =0.78 m s-1, foR =21.0%, exhQ =120 m3 h-1 at the downwind exhaust. X=4.25 was at the

upwind side edge of the pit, and X=5.75 was at the downwind side edge of the pit. For the

downwind side, both mean and turbulent flux approached 0 g s-1 m-2. For upwind side, the

vertical mean flux was negative and moved downward into the pit. The average mean flux

was -11.09 g s-1 m-2. The vertical turbulent flux was positive and moved upward to wind

tunnel space. The average turbulent flux was 30.25 g s-1 m-2. Fig. 6.9 indicated that the total

vertical flux was positive by integrating mF and tF along X-axis.

Fig. 6.9 – Horizontal distributions of the vertical mean and turbulent flux of CO2 through the top surface of the centre slot (see Fig.4). The simulation was performed under the condition

of HU =0.78 m s-1, foR =21.0%, exhQ =120 m3 h-1 and downwind exhaust

6.4. Discussion

6.4.1. The simplification of computational domains

Simplifications of detailed geometries are needed to simulate reality. The purpose of

simplification could be: (1) the difficulty in including some parts in simulations, for example,

Page 119: NEERENGI - Aarhus Universitet

105

it is hard to use detailed locations and shapes of animals in a livestock house; (2) to reduce

number of cells, for example, the narrow openings of slatted floor cause a large number of

cells, which would be reduced by treating slatted floor as porous media (Bjerg et al, 2008a);

(3) to simplify boundary conditions, for example, the domain 1 in this work was intended to

avoid modelling the fans of the wind tunnel directly. But the accuracy of simplification of

CFD domain was usually not evaluated in literature. By comparison of two computational

domains, a significant effect of domain was found. Domain 2 failed to predict removal ratios.

The reason could be as follow. In domain 2, the wind table was extended to whole wind

tunnel. Hence, the airflow above the wind table was parallel to slats and the vertical velocity

magnitude was nearly zero everywhere. The free stream in the wind tunnel confined the

airflow inside the pit. Therefore, more than 90% gases were exhausted by the pit ventilation

system in all cases. However, in domain 1, there was space between side-edge of wind table

and sidewall of wind tunnel. Vertical air circulation was observed above wind table. The

vertical vortex can bring more contaminant into the space of wind tunnel. Thus domain 1

predicted removal ratios more precise than domain 2.

6.4.2. Velocity fields and effects of different RANS turbulence models

The discrepancy between the simulated and measured velocities can be due to the measuring

instrument. An ultrasonic anemometer was used to measure the air velocities. The sensor

head/measurement volume was comparatively large to the headspace of the slurry pit. It

covered a volume larger than 0.1m×0.1m×0.1m. The measured velocity was actually the

averaged value in the air space covered. The velocity gradient (Fig. 6.5b) was much steeper

on the downwind side according to the simulation. The averaged value by the ultrasonic

anemometer in some regions of the downwind side is very different from the exact point

value achieved by CFD. This can be the reason that the discrepancy of velocities at

downwind side was larger than that on upwind side.

Another reason of the discrepancy between CFD and measurement results could be the

defined air velocity boundary condition generated by the fan. Since detailed information of

the wind tunnel fan was not available, an assumption of a fixed velocity at wind tunnel inlet

based on the calibrations by the measured velocity profile above wind table was applied (Fig.

6.2). The fixed velocity might not well describe the boundary condition generated by the

fans.

The third reason might be the inability to apply k and other RANS turbulence

models on the cavity flow. In this study, the cavity flow was featured with strong streamline

curvatures and separation. Isaev et al. (2006) reported that the k turbulence models

sharply overestimated the eddy viscosity in the primary vortex zone and underestimated the

separated flow intensity. A correction for the curvature of streamlines in the expression of the

Page 120: NEERENGI - Aarhus Universitet

106

turbulent viscosity definition was not proposed in the k turbulence models. The

three k turbulence models underestimated velocities near upwind wall and overestimated

velocities near downwind wall of pit. Transition SST kw highly overestimated velocities

near exhaust openings and resulted in a largest prediction discrepancy on removal ratio

compared with other RANS models. Transition SST wk was used to consider low-Re effect

inside the pit but it did not improve the simulated results. Results from RSM agreed best with

the measurements. Although it took much more iterations for a converged solution, RSM was

most promising according to this study.

6.4.3. Removal ratios predicted by CFD

Removal ratios ( Rr ) based on CFD simulations were very close to measured values in most

cases. It was feasible to use RSM to assess the ability of a partial pit ventilation system to

reduce emission under slatted floor.

The same conclusions can be drawn from both measurements and simulations: (1) the

changes of Rr following the variation of air velocity above the floor was similar to

simulations and measurements for HU <0.78 m s-1 at the downwind exhaust or all upwind

exhaust cases; (2) smaller slatted floor opening can lead to higher removal

capacity/efficiency; (3) upwind exhaust was more efficient than downwind exhaust.

Differences between measured and simulated results were also found. In the

measurements, the mean Rr was 83.1% for all cases upwind exhaust cases and some

downwind exhaust cases. There was no clear velocity dependency (Wu et al., 2012).

However, when HU increased from 0.78 to 1.94 m s-1, the simulated Rr decreased from

87.7% to 77.2% for the upwind exhaust and showed the same velocity dependency in the

cases of downwind exhaust. For the upwind exhaust in the measurements, higher ventilation

rates (120 m3 h-1) did not always remove more emission. The reason for this case can be that

the airflow patterns were different for ventilation rates 120 and 60 m3 h-1 (Wu et al., 2012).

But in the simulations, higher ventilation rates (120 m3 h-1) always resulted in higher removal

ratio. The airflow patterns (Fig. 6.7b) at the two ventilation rates were similar to each other

from the simulations.

The differences between measured and simulated values could arise for several reasons.

The first reason can be the difference of turbulence diffusion in the measurement and in the

simulation. Both the mean flow field and the turbulence transportation controlled the gas

dispersion through the slatted floor. Fig. 6.9 shows that the CO2 transportation to free stream

above the slatted floor was dominated by the turbulent diffusion, which was consistent with

the studies on scalar transport in street canyons (Lee and Park, 1994; Baik and Kim, 1999).

The RSM turbulence models probably had difficulty to estimate turbulence diffusion

accurately in the slatted slots.

Page 121: NEERENGI - Aarhus Universitet

107

Unsteady mass transport phenomena can be another reason. Wu et al. (2012) reported

that the standard deviation of CO2 concentration was very large compared with its mean

value. It indicates that the transportation of CO2 was rather unsteady. It was also consistent

with the experiment and computational studies of Chang and Meroney (2003), which

revealed that the dispersion of gaseous pollutants within the cavity of street canyons was

essentially unsteady. Johnson and Hunter (1998) also pointed out that transient flow was a

very important factor for modelling passive pollutant concentration fields from the wind

tunnel measurements. The nature of RANS models is a steady state methodology. Therefore,

it was limited to take transient turbulent transportation of CO2 into account.

In order to validate the explanations, further measurements should be conducted to

depict the airflow patterns inside the pit, which may require advanced measurement setups

and instrumentations. An unsteady CFD methodology such as large eddy simulation could be

used to assess the effect of unsteady transportation of scalar.

6.5. Conclusion

RSM (Reynolds Stress Models) was found to be the best of the used RANS turbulence

models.

RSM can be used to predict the removal capability of a partial pit ventilation system to

reduce emission under slatted floor.

Turbulence diffusion dominated the mass transfer from pit headspace into free stream

above the floor.

Future research should focus on depicting the airflow patterns inside the pit by using

scale model under well controlled laboratory conditions, and generating benchmark data to

validate and improve CFD methods. Unsteady CFD simulation using large eddy simulation

could be conducted to make more precise predictions of removal capability of the partial pit

ventilation system by considering the unsteady phenomena of gas emission.

References

Baik, J.J., Kim, J.J., 1998. A numerical study of flow and pollutant dispersion characteristics

in urban street canyons. Journal of applied meteorology 38, 1576-1589.

Baik, J.J., Kim, J.J., 2002. On the escape of pollutants from urban street canyons.

Atmospheric Environment 36, 527-536.

Bartzanas, T., Boulard T., Kittas, C., 2002. Numerical simulation of the airflow and

temperature distribution in a tunnel greenhouse equipped with insect-proof screen in the

openings. Computers and Electronics in Agriculture 34, 207-221.

Bjerg B., Andersen M., 2010. Numerical simulation of a pit exhausts system for reduction of

ammonia emission from a naturally ventilated cattle building. XVIIth world congress of

Page 122: NEERENGI - Aarhus Universitet

108

the international commission of agricultural and biosystems engineering (CIGR).

Quebec City, Canada.

Bjerg B., Zhang, G., Kai, P., 2008a. CFD investigations of a partly pit ventilation system as

method to reduce ammonia emission from pig production units. The Eighth ASABE

International Livestock Environment Symposium (ILES Ⅷ)

Bjerg B., Zhang, G., Kai, P., 2008b. Porous media as boundary condition for air inlet, slatted

floor and animal occupied zone in numerical simulation of airflow in a pig unit.

AgEng2008 International Conference on Agricultural Engineering, Hersonissos, Crete-

Greece.

Castro, I.P., Apsley, D.D., 1997. Flow and dispersion over topology: a comparison between

numerical and laboratory data for two dimensional flows. Atmospheric Environment

31, 839-850.

Chang, C.H., Meroney, R.N., 2003. Concentration and flow distributions in urban street

canyons: wind tunnel and computational data. Journal of wind engineering and

industrial aerodynamics 91, 1141-1154.

Choudhury, D., 1993. Introduction to the Renormalization Group Method and Turbulence

Modeling. Fluent Inc. Technical Memorandum TM-107.

Davidson, L. 1997. An Introduction to Turbulence models. Dept. of Thermo and Fluid

Dynamics, Chalmers University of Technology, Gothenburg, Rept. 97/2.

Hanjalic, K. 1999. Second-Moment Turbulence Closures for CFD: Needs and Prospects. Int.

J. Heat and Fluid Flow 12, 667-697.

Huang, H., Akutsu, Y., Arai, M., Tamura, M., 2000. A two-dimensional air quality model in

an urban street canyon: evaluation and sensitivity analysis. Atmospheric Environment

34, 689-698.

Isaev, S.A., Baranov, P.A., Kudryavtsev, N.A., Lysenko, D.A. and Sachov, A.E. U., 2005.

Complex analysis of turbulence models, algorithms, and grid structures at the

computation of recirculating flow in a cavity by means of VP2/3 and Fluent packages.

Part 2. estimation of models adequacy, Thermophysics and Aeromechanics 13 (4), 55-

65.

Johnson, G.T., Hunter, L.J., 1998. Urban wind flows: wind tunnel and numerical simulations

—a preliminary comparison. Environmental Modelling and Software 13, 279–286.

Kim, J.J., Baik, J.J., 2004. Effects of inflow turbulence intensity on flow and pollutant

dispersion in an urban street canyon. Journal of Wind Engineering and Industrial

Aerodynamics 91, 309-329.

Launder, B.E., Spalding, D.B., 1974. Numerical computation of turbulent flows. Computers

and Mathematics in Applied Mechanics and Engineering 3, 269-289.

Page 123: NEERENGI - Aarhus Universitet

109

Lee, I.Y., and Park, H.M., 1994. Parameterization of the pollutant transport and dispersion in

urban street canyons. Atmospheric Enviroment 28, 2343-2349.

Li, X.X., Liu, C.H., Leung, D. Y.C., Lam, K.M., 2006. Recent progress in CFD modelling of

wind field and pollutant transport in street canyons. Atmospheric Environment 40,

5640-5658.

Menter, F.R. 1994. Two-equation Eddy-viscosity Turbulence Models for Engineering

Applications. AIAA J. 32, 1598-1605

Rong L., Nielson, P. V., Zhang, G., 2010. Experimental and numerical study on effects of

airflow and aqueous ammonium solution temperature on ammonia mass transfer

coefficient. J.Air & Waste Manage. Assoc. 60, 419-428.

Sagrado, A. P. G., Beeck, J., Rambaud, P., Olivari, D., 2002. Numerical and experimental

modelling of pollutant dispersion in a street canyon. Journal of Wind Engineering and

Industrial Aerodynamics 90, 321-339.

Saha CK; Wu W; Zhang G; Bjerg B (2011). Assessing effect of wind tunnel sizes on air

velocity and concentration boundary layers and on ammonia emission estimation using

computational fluid dynamics (CFD). Computers and Electronics in Agriculture, 78(1):

49-60

Sahm, P., Louka, P., Ketzel, M., Guilloteau, E., Sini, J. F., 2002. Intercomparison of

numerical urban dispersion models – part1: street canyon and single building

configurations. Water, Air, and Soil Pollution: Focus 2, 587-601.

Shih, T.-H., Liou, W. W., Shabbir, A., Yang, Z., Zhu, J. 1995. A New k Eddy-Viscosity

Model for High Reynolds Number Turbulent Flows - Model Development and

Validation. Computers Fluids 24(3), 227-238.

Sini, J. F., Anquetin, S., Mestayer, P. G., 1996. Pollutant despersion and thermal effects in

urban street canyons. Atmospheric Environment 95, 1352-2310.

Tsai, M.Y., Chen, K.S., 2004. Measurements and three dimensional modelling of air

pollutant dispersion in an urban street canyon. Atmospheric Environment 38, 5911-

5924.

Wu, W., Kai, P., Zhang, G., Bjerg, B., Nielsen, P. V., 2011. Assessment of a partial pit

ventilation system to reduce emission under slatted floor – part1: Scale model study.

Manuscript submitted to Computers and Electronics in Agriculture.

Zhai, Z., Zhang, Z., Zhang, W., Chen. Q., 2007. Evaluation of various turbulence models in

predicting airflow and turbulence in enclosed environments by CFD: Part-1: summary

of prevent turbulence models. HVAC & R research 13(6), 1-21.

Page 124: NEERENGI - Aarhus Universitet

110

Chapter 7

Large eddy simulation of airflow and pollutant transport under slatted floor of a slurry pit model

Paper VI:

Wu, W., Zong, C., Zhang, G., 2012. Large eddy simulation of airflow and pollutant transport

under slatted floor of a slurry pit model. (Submitted to a peer review journal)

Page 125: NEERENGI - Aarhus Universitet

111

Abstract

For dairy cattle buildings with slatted floor systems, about 40% of the ammonia emission originates

from the slurry pit. In order to find a solution to abate this part of emission, a better understanding

of the ammonia transportation from the pit to the room space is crucial. Large eddy simulation

(LES) was adopted to investigate the transportation of airflow and ammonia under slatted floor. To

tackle the involvement of the slatted floor, two approaches were proposed: modelling slatted floors

directly with geometrical details (LESD) and treating them as porous media (LESP). The main

purpose of this work was to study the potential of using porous media to model the slatted floor. The

LES results were validated by the air velocities measured using a LDA (Laser Doppler

anemometer) in a 1:8 scale pit model placed in a wind tunnel. The results showed that LESP was

able to estimate the mean air velocities and turbulence kinetic energy in the core of the pit

headspace; but it cannot satisfactorily predict the mean air velocities and turbulence kinetic energy

in the space next to the upwind wall. Apparent vertical air motion in the top surface of the slot was

observed for LESD results. There was not such trend found for LESP results. Both the air velocity

and NH3 mass fraction fluctuated weaker for LESP results. By spectral analysis, LESP was able to

capture the entire power spectrum compared with LESD. A dominant Strouhal number 0.23 was

found for LESD results but no dominant strouhal number was found for LESP results. The emission

rate and total mass of NH3 in the pit headspace calculated by LESD was double of those calculated

by LESP. Pollutants were confined in the headspace for longer time by means of using LESP than

using LESD. For both LESD and LESP, turbulence transportation was the dominant removal

mechanism to transport pollutants from the headspace to the free stream.

Key words: CFD; LES; porous media; wind tunnel; Strouhal number

Page 126: NEERENGI - Aarhus Universitet

112

7.1. Introduction

Dairy cow buildings constitute one of the largest sources of ammonia emissions among various

animal husbandry operations in Denmark (Pederson, 2006). Solutions to curtail the dairy cow

emissions include changes in housing and floor design. Slatted floor with scrapers on the floor

surface as well as channel scraper show the potential to keep ammonia emissions at a low level

(Zhang et al., 2005; Wu et al., 2012a). For a cattle building with slatted floors, about 60% of the

total ammonia emission originates from floor surface, whereas 40% is released from the slurry pit

(Braam et al., 1997). In order to find a solution to abate the portion of emission coming from the

slurry pit and improve the design of the floor system, a better understanding of the ammonia

transportation mechanism from the pit to the room space is necessary. So far, some studies

regarding airflow patterns and pollutant dispersion in dairy cow buildings have been limited to the

space above the slatted floor (Norton et al., 2010a; Norton et al., 2010b; Wu et al., 2012b).

However, the knowledge of the characteristics of the airflow and mass transport under the slatted

floor is still missing, although it is crucial to estimate the ammonia emission from the slurry pit

correctly. Wu et al. (2012c) applied CFD to study the airflow characteristics under slatted floor in a

1:2 scale pit model, in which the slatted floor was simulated in geometrical detail. However, the

measured air velocities were limited due to the spatial resolution of the sensors and the

experimental facilities (Wu et al., 2012d). Considering that obtaining the characteristics of the air

motion under slatted floor in a full scale livestock building is the final research goal, modelling of

slatted floor in reality becomes the main concern. The slot width in a real cattle building is about

0.02 m, while the shortest building dimension is generally longer than several metres. The small

ratio of slot width and building dimension prevents a direct modelling of the geometrical details.

Therefore, slatted floor is usually handled as porous media (Sun et al., 2004; Bjerg et al., 2008a;

Bjerg et al., 2008b). However, up to date, the information on the difference that might exist between

simulations using geometrical details and porous media cannot be found in literature. The

uncertainty of using porous media to replace the slatted floor above the slurry pit in simulations

should be investigated.

The crucial prerequisite of an investigation on the difference between modelling slatted floors

with geometrical details and treating them as porous media is to accurately simulate the flow

passing by the pit headspace - a cubic cavity. The phenomenon is known to be difficult to model

due to the flow separation. The research of similar cavity flow can be found in the area of modelling

airflow in street canyons (Vardoulakis et al., 2003). Reynolds-averaged Navier-Stokes (RANS)

models are the most commonly adopted turbulence models in calculating street canyon wind flow.

The RANS models ever employed in street canyon field are the standard k model (Johnson and

Hunter, 1998; Baik and Kim, 1999; Baik and Kim, 2002; Kim and Baik, 2003; Neofytou et al.,

2006) and its variants, RNG k model (Tsai and Chen, 2004) as well as realizable k model

Page 127: NEERENGI - Aarhus Universitet

113

(Sagrado et al., 2002). Despite of the widely utilization of k models, the main deficiency

pointed out by Solazzo et al. (2009) is the underestimation of the turbulence kinetic energy and air

velocities within some parts of the cavity. The experimental and computational studies of Chang

and Meroney (2003) revealed that the dispersion of gaseous pollutants within the street cavity was

essentially unsteady. Johnson and Hunter (1998) also reported that the transient flow was a very

important factor in quantifying the gas dispersion from street canyon. Hence, the precise prediction

of gas dispersion from a cavity may not be reached by assuming a steady state process. Due to the

weakness of RANS models, large eddy simulation (LES) was used to examine the flow features in

street canyons by Liu et al. (2005) and the LES results agreed reasonably well with wind tunnel

measurements. Walton and Chen (2002) compared the accuracy of k and LES turbulence

closure schemes against experimental results and found that the LES results exhibited the best

agreement with measured results.

The cavity flow in this study is much more complicated than the above mentioned studies on

street canyon due to the involvement of a porous layer (the slatted floor). The main purpose of this

work is to use Large Eddy Simulation (LES) to investigate the difference of airflow and pollutant

transportation in the pit headspace between modelling slatted floor directly (LESD) and modelling

slatted floor as porous media (LESP).

7.2. Materials and methods

This section will start with an introduction to the wind tunnel measurements used to validate the

large eddy simulation. It follows with a description of the development of the CFD model and

theories of LES. At the end of this section, the methods to analyze the simulation results will be

presented.

7.2.1. Measurements for model validation

7.2.1.1. Experimental apparatus

The experiment was carried out in a wind tunnel at the air physics lab, Aarhus University,

Denmark. The wind tunnel (Fig. 7.1 a) was 3.67 m long with cross section area of W×H = 0.35

m×0.35 m. At working section of the wind tunnel there was a 0.80 m long transparent glass window

for velocity measurement using a Laser Doppler anemometer (LDA). Smoke was injected at the

wind tunnel inlet to provide the seeding for LDA to measure air velocity. A ventilator (Type CK

200 B CBU, Lindab A/S, Denmark) was installed at the tunnel outlet to drive the air motion through

the tunnel.

A 1:8 scale pit model (Fig. 7.1b) with Lp×Wp ×Hp = 0.35 m×0.35 m×0.09 m was constructed

at the working section. The top of the pit model was covered with 17 slats, which were orientated

Page 128: NEERENGI - Aarhus Universitet

114

parallel to the flow direction. The dimension of the slats is shown in Fig. 7.1b. The top of the slatted

floor was at the same level with the surface of the tunnel floor.

(a)

(b)

(c)

Fig. 7.1 Schematic of (a) the wind tunnel; (b) the pit model and the slats; (c) the velocity

measurement locations (L1-L7). All dimensions are in m.

7.2.1.2. Air velocity and turbulence measurement

Air velocity and turbulence were recorded at 7 vertical heights (0.022, 0.027, 0.032, 0.037, 0.042,

0.047 and 0.057 m) from the bottom surface of the pit and at 7 locations L1-L7 (Fig. 7.1c) along the

longitudinal direction of the pit. These 7 locations were under a slot in the middle plane of the wind

tunnel and the distance between each other was 0.05 m. The first location was 0.031 m away from

the windward wall of the pit.

Page 129: NEERENGI - Aarhus Universitet

115

A 2-D LDA (Type 58N40-FVA enhanced, DANTEC Dynamics, Skovlunde, Denmark) was

used to measure the air velocities and turbulence quantities. The work principle of LDA was

clarified by Saha et al. (2010). The recording time for each position was 2.5 min.

7.2.2. CFD modelling

7.2.2.1. Numerical representation of the pit model with slatted floor

Fig. 7.2a shows that a slot bounded with two half slats forms a rectilinear geometry in the cross

direction of the flow. The periodical occurrence of slats distribution enables a slot with two adjacent

half-slats to represent the entire slatted floor system. It was observed in the experiment that the

velocity gradient was zero at 0.5H height of the wind tunnel. Therefore, the half height of the wind

tunnel space was used in the CFD model. Fig. 7.2b illustrates the simplified geometry for the CFD

model.

(a) (b)

Fig. 7.2 (a) The CFD representation of the pit model geometry with slatted floor; (b) CFD domain,

mesh and boundary conditions

7.2.2.2. Boundary conditions and mesh

The computational domain with mesh and boundary conditions is shown in Fig. 7.2b. The height of

the domain was 0.5H. The free surface layer extended 3Hp and 5Hp in upstream and downstream

directions, respectively. The top surface of the free stream was defined as zero gradient boundary

condition. The ventilator created an air speed 1.41±0.11 m s-1 at 0.5H height, which was used as the

inlet air velocity (U). The associated Reynolds number was 8.69×103 based on the inlet velocity U

Page 130: NEERENGI - Aarhus Universitet

116

and the height of the pit Hp. Pressure outlet was prescribed as the outlet boundary condition.

Symmetry planes were imposed on the two domain sides along the streamline direction, which

allowed a slot and two half slats to translate themselves periodically towards the direction

perpendicular to the free streamlines. No-slip wall was imposed on the solid surfaces. As above

mentioned objectives, the slatted floor was modelled directly in one case and was handled as porous

media in another case. In the case of simulating slatted floor directly, the volume occupied by slats

was defined as solid material. Otherwise, the volume was defined as air, the identical material with

the whole fluid domain. A resistance was added to the cell zones representing the space of the two

half slats and the slot. The resistance thus generated the pressure drop when air went though the cell

zones. A concrete description of modelling flow through porous media can be found in the work of

Wu et al. (2012c). The resistance coefficient of the porous zone was taken as 10000 m-2 for viscous

resistance and 50 m-1 for inertial resistance. The estimation of the resistance coefficients was based

on the methodology proposed by Bjerg et al. (2008a).

Three mesh systems with 148876, 253140 and 454458 cells were constructed to test grid

independence. Fig. 7.3 shows the horizontal air velocities with respect to heights at location L5. The

medium mesh predicted very similar air velocities with fine mesh. Therefore, the medium mesh

with 253140 cells was used for further calculations.

Fig. 7.3 Mesh convergence study of the velocities at L5: dotted line – coarse mesh; solid line –

medium mesh; dashed line – fine mesh.

Since the simulations were activated in a transient mode, the time step should be defined

according to Courant-Friedrichs-Lewy condition,

tConsz

tu

y

tu

x

tu zyx tan

(7.1)

where, x, y, z are coordinates; ux, uy, uz are velocity scales; x , y , z are grid spaces in three

coordinates, respectively; t is time step. The smallest grid was 2×10-4 m for all the three

coordinates and the free stream velocity U were set as the reference velocity scale for the three

dimensions. The time step was 1×10-4 s. The height of the pit headspace Hp was taken as the length

Page 131: NEERENGI - Aarhus Universitet

117

scale. The time scale T was calculated as Hp/U, which was 0.064 s. The simulations were run 50 T

for the turbulent flow to achieve quasi-steady state. Another 50T was computed for data sampling.

7.2.2.3. Turbulence modelling

LES approach computes the larger eddies for each problem with a time dependent solution. The

universal behaviours of small eddies are captured with a compact model. Instead of time averaging,

LES uses a spatial filtering operation to separate the larger and smaller eddies. The method starts

with the selection of a filtering function and a certain cut-off width. Box filter used in the finite

volume implementations can be written as follow

0

/1),',(

3

xxG 2/'

2/'

xx

xx

(7.2)

where, ),',( xxG is a filter function; is filter cut-off width. In the next step spatial filtering

operation is performed on the time dependent flow equations. The spatial filtering operation is

defined as

'3

'2

'1),'(),',(),( dxdxdxtxxxGtx

(7.3)

where, ),( tx is filtered function; ),'( tx is original function. Applying the filter operation (3) to the

unsteady and incompressible continuity and Navier-Stokes equations yields the LES equations,

0

i

i

x

u

(7.4)

jj

i

j

ij

iji

j

i

xx

u

xx

puu

xt

u

2

(7.5)

where, iu , ju are the larger scale velocities in the i, j direction; p is larger scale pressure. is

dynamic viscosity. After spatial filtering operation, information related to the smaller eddies which

is filtered out is described by means of sub-grid-scale stresses (SGS), denoted as ij in equation (5).

jijiij uuuu (7.6)

The SGS stresses is computed by employing Boussinesq hypothesis,

Page 132: NEERENGI - Aarhus Universitet

118

ijtijiiij S 23

1

(7.7)

in which ij is Kronecker’s delta; t is the SGS turbulent viscosity; the rate of strain tensor ijS is

defined by

i

j

j

iij x

u

x

uS

2

1

(7.8)

In this study, t is provided by Smagorinsky-Lilly model,

SLst2 (7.9)

where, Ls is the mixing length for SGS. It is determined by

),min( ss CdL (7.10)

where, is the von Karman constant, d is the distance to the closest wall, Cs is the Smagorinsky

constant (0.1 is used in this work).

7.2.2.4. Modelling pollutant transportation

Ammonia is considered as the pollutant in this study. Applying the filter operation (7.3) to the

species transportation equation gives the mass equation for LES,

iii

cii

i xx

cD

xcu

xt

c

2

,

(7.11)

where, c is the filtered ammonia fraction; ci , is SGS flux term

cucu iici , (7.12)

which is modelled using the same methodology for SGS stresses. D is the mass diffusivity and can

be estimated by (Sommer et al., 2006)

23/13/1

2/13

75.17

)9.141.20(

)/1/1(*10

p

MMTD airNH

(7.13)

where, M is the molecular weight; p is the atmospheric pressure, atm; T is the air temperature, K;

the meaning of all the constants is explained by Sommer et al. (2006).

7.2.3. Strouhal number

The difference of simulating slatted floor directly and indirectly (as porous media) may be

illustrated by the vortex oscillation in the headspace of the pit. The oscillating flow under the slatted

floor can be described by Strouhal number (St),

Page 133: NEERENGI - Aarhus Universitet

119

U

fHSt p

(7.14)

where, f is the frequency of the vortex vibration in the headspace of the pit model. For large St

(order of 1), viscosity dominates the flow; for low St, the quasi steady state portion of the fluid

movement dominates the oscillation.

7.2.4. Retention time

Liu et al. (2005) defined a time scale of pollutant residing in a street canyon as the pollutant

retention time. The concept of retention time is adopted in this study. The aim is to identify the

consistency of the time scale of pollutant residing under slatted floor while the slatted floor is

simulated directly and indirectly (as porous media). The retention time ( ) is determined by

Q

(7.15)

where, Q is the pollutant emission rate, which can be obtained by integrating the pollutant in the

outlet of the free stream; Θ signifies the total mass of the pollutant confined under the slatted floor,

which can be calculated using

dc (7.16)

where, Ω is the volume of the headspace of the pit.

7.2.5. Vertical mass flux induced by turbulent diffusion and mean flow

Baik and Kim (2002) proposed two formulas CW and c'w' to calculate vertical mass flux induced by

mean flow and turbulence, respectively. c' and w' are the deviations from the mean NH3 mass

fraction C and mean vertical velocity W, respectively at the slatted floor level. Wu et al. (2012c)

employed CW and c'w' to study the transportation mechanism of pollutant through the slatted floor.

c'w' may help to find the difference of transportation mechanism through a real slot (simulating

slatted floor directly) and through a porous zone (simulating slatted floor as porous media).

7.3. Results and discussion

7.3.1. Model validation

7.3.1.1. Comparison of air velocity profiles

Fig. 7.4 shows the vertical profiles of mean air velocity in the pit headspace. At L1 and L2 near the

upwind wall, LESD results achieved very good agreements with measured results both on the shape

of the velocity profile and on each velocity component (Ux, Uz), while LESP cannot predict correct

mean air velocities at these

Page 134: NEERENGI - Aarhus Universitet

120

Page 135: NEERENGI - Aarhus Universitet

121

(a) (b) Fig. 7.4 Vertical profiles of mean air velocity in the pit headspace: triangle – measured values; solid

line – simulated values by modeling SF directly; dashed line – simulated values by treating SF as

porous media. (a) Ux – horizontal velocity; (b)Uz – vertical velocity.

two locations. When the flow reached the core of the headspace (L3-L5), both the simulations were

able to predict mean air velocities in line with the measured values and LESP results showed a little

better agreement with measurement results compared with LESD results. For the vertical air

velocity (Uz), the profiles obtained by LESD had four peaks at -0.0025, -0.01, -0.02, -0.03 m, but

the profiles calculated by LESP were more constant. Next to the downwind wall (L6 to L7), the

Page 136: NEERENGI - Aarhus Universitet

122

simulated Ux, Uz using LESD and LESP fitted well with the measured values. But the profile shapes

were quite different. Generally, the LESD showed better performance than the LESP.

7.3.1.2. Turbulence modelling

Fig. 7.5 shows the vertical profiles of turbulence kinetic energy in the pit headspace. A large

discrepancy was found between the turbulence kinetic energy predicted by LES and measured

values at L1. At L2 to L7, both the LES results can agree well with the measured results. The

profile pattern from LESD and LESP was similar except that the kinetic energy calculated by LESD

was almost 4 times of that calculated by LESP in the slot (around -0.01 m).

Page 137: NEERENGI - Aarhus Universitet

123

Fig.5 Vertical profiles of turbulent kinetic energy in the pit headspace: triangle – measured values; solid line – simulated values by modeling SF directly; dashed line – simulated values by treating SF

as porous media.

7.3.2. Flow field and NH3 distribution in the headspace

Fig. 7.6 shows the time-averaged airflow patterns, velocity magnitude and NH3 mass fraction in the

pit headspace obtained by LESD and LESP. Both modelling methods predicted 3 vortices in the

headspace. Via treating the slatted floor as porous media, the 3 vortices shifted slightly to the

upwind direction. For LESD results, a big circulation started to be formed by the reflection of the

downwind wall and was extended to the upwind wall. For LESP results, a similar eddy started to

circulate from the downwind wall but did not reach the downwind wall and was ended in the middle

section of x coordinate. Air velocity in the slot was about 0.5 to 1 m s-1, while it was smaller than

0.2 m s-1 in the porous zone. The higher air velocity in the slot enabled the air to attach the bottom

surface of the pit at a higher speed 0.3 m s-1 and had more energy to travel long distance to the

upwind side. LESP results showed that the process of the attachment to bottom surface of pit was

weaker compared with LESD results. This may explain the difference of the big circulation

obtained by the two different methods. Another difference of the airflow patterns between the two

simulation methods should be noticed. For LESD results, the air on the top surface of slot had

significant vertical movement and can flow out of the cavity. However, for LESP results, there was

no significant vertical air movement on the interface of the porous layer and the tunnel space.

High concentration of NH3 was found in the upwind side of the pit headspace. The NH3 mass

fraction obtained by LESD was higher than that calculated by LESP.

7.3.3. Velocity and NH3 fluctuation in the headspace

Fig. 7.7 shows the time history of air velocity and NH3 fraction at P0. The air velocities obtained by

LESD were more fluctuated and had a higher mean value of 0.28 m s-1 while the LESP results were

less oscillated and had a lower mean air velocity of 0.12m s-1. The NH3 fraction displayed same

fluctuation feature with the air velocities at P0 for two the simulation methods. But the oscillation

Page 138: NEERENGI - Aarhus Universitet

124

of mass transfer was slower than the oscillation of flow. During 50T, flow finished about 12 periods

while the mass transfer underwent only about 3 circles.

(a) (b)

Fig. 7.6 Airflow patterns, velocity magnitude, ammonia mass fraction in the pit headspace: (a)

modeling SF directly; (b) treating SF as porous media.

Page 139: NEERENGI - Aarhus Universitet

125

(a) (b) Fig. 7.7 Time history of ensemble-averaged variables at point P0: (a) air velocity magnitude and

(b) mass faction of ammonia in the pit headspace. Solid and dashed lines denote results from modeling SF directly and treating SF as porous media, respectively.

Fig. 7.8 (a) shows the power spectral density of air velocity with respect to frequency at P0.

The overlap of the two spectrums suggested that LESP method can capture the entire power

spectrum of vortex oscillation in the pit headspace. Fig. 7.8 (b) shows the power spectral density

against the Strouhal number. A dominant St existed for the LESD results and was approximated

0.23. However, the dominant St did not occur for the LESP results.

(a) (b) Fig.7.8 Spectral analysis of time history of air velocity magnitude on point P0: (a) power spectral

density with respect to frequency; (b) power spectral density with respect to Strouhal number. Solid

and dashed lines denote results from modeling SF directly and treating SF as porous media,

respectively.

The higher fluctuation in LESD may be explained as follow. The presence of slats in the

simulation introduced turbulence kinetic energy production inside the slot. This can be proved by

the higher kinetic energy in the slot at L1to L7 calculated by LESD compared to that calculated by

Page 140: NEERENGI - Aarhus Universitet

126

LESP (Fig. 7.5). After the slatted floor was handled as porous media, the turbulence kinetic energy

generated by slats cannot be recovered in the zone of slatted floor. Therefore, the turbulence kinetic

energy generated by slats caused flow in the pit headspace more unstable and fluctuated. The

turbulence kinetic energy generated by slats and the flow separation of the shear layer in the vicinity

of slats might be attributed to the instabilities and vortex oscillation at P0 in the LESD simulation.

7.3.4. Retention time

Retention time defined by equation (7.15) was used to investigate the time scale of the NH3

transport in the pit headspace in order to study the difference of the two simulation methods. The

NH3 emission rate, total NH3 emission confined in the pit headspace and retention time are given in

Table 7.1. The emission rate and total mass confined in the headspace of NH3 calculated by LESD

was double of those from LESP. The vertical air motion (Fig. 7.6) may cause more pollutants to

escape from the headspace by means of LESD simulation. However, the difference of the retention

time for the two simulation methods was 0.29 s. Given that the reference time scale was 0.064 s, the

dimensionless difference of retention time was 4.5T. The NH3 was confined in the cavity for 4.5T

longer by using LESP than by using LESD.

Table 7.1 Emission rate, Total mass in the pit headspace and retention time of NH3

Modelling method Emission rate (mg s-1) Total mass in the pit headspace (mg) Retention time (s) LESD 34.3 215 6.27 LESP 16.8 110 6.56

7.3.5. The pollutant removal mechanism for the two simulation methods

Fig. 7.9 shows the vertical NH3 flux transported by mean flow Uz and turbulence u’z in LESD and

LESP. For both of the simulation results, more than 87% of the pollutants were removed by

turbulence transportation (87.9% in LESD, 87.4% in LESP). In the downwind side, the mean flow

even transported the pollutants back to the cavity. The negative flux was -0.65 and -1.88 mg s-1 for

LESD and LESP results, respectively. The maximum removal of pollutants happed at position x=-

0.05 m (0.05 m away from the cavity centre and in the upwind side) for both simulation methods.

The maximum transportation (84.3 mg s-1 for LESD and 59.1 mg s-1 for LESP) was driven by

turbulence. This agrees well with other research reports and further confirms that the pollutant

removal from a cubical cavity was mainly due to turbulence diffusion (Lee and Park, 1994; Baik

and Kim, 1998; Wu et al., 2012d).

7.3.6. Further discussion and future work

Through the examination of the performance of two numerical methods to model the airflow and

pollutant transportation under slatted floor, LESP seemed able to estimate the mean air velocities

and turbulence kinetic energy in the core of the pit headspace. However, differences between results

of LESD and LESP also existed. The major difference could be due to the inability of LESP to

Page 141: NEERENGI - Aarhus Universitet

127

include the turbulent kinetic energy generated by the presence of slats. This part of kinetic energy

affected the vertex oscillation and the pollutant removal through the slots. Due to the strong

oscillation led by slats, more pollutant had the chance to escape the cavity (Fig.9). To make the

utilization of porous media more efficient to represent slatted floor, not only a pressure drop was

added to the Navier-Stoke (N-S) equation but also a correction to the extra energy production in the

slots should be well considered.

Fig.7.9 Vertical ammonia flux transported by Uz and u’

z: Solid and dashed lines denote results from

modeling SF directly and treating SF as porous media, respectively; Thick and thin lines mark

ammonia transportation by Uz and u’, respectively

The time scale for pollutants to reside in the cavity was found inconsistent for LESD and

LESP. One reason might be due to the change of the length scale when porous media was used

instead of modelling slatted floor directly. The narrow slot may decompose the large eddy from the

free stream to small eddies. The process can cause turbulent kinetic energy cascade and dissipation.

When the flow went through the slots, the length scale was changed again. However, in the current

LES, the variation of the length scale was not considered in the N-S equation.

This work only considers the case when the slats were oriented parallel to the flow direction.

To reduce uncertainties of using porous median, different orientations of slats such as perpendicular

to the flow direction should be investigated in the future. Correction of turbulence kinetic energy

production in the porous zone will also be the perspective work.

7.4. Conclusions

The main objective of this work was to evaluate the performance of using porous media to simulate

slatted floor. Two numerical methods, modelling slatted floors directly with geometrical details

(LESD) and treating them as porous media (LESP), were proposed. Main conclusions can be drawn

from the results:

Page 142: NEERENGI - Aarhus Universitet

128

(1) LESP was able to estimate the mean air velocities and turbulence kinetic energy in the core

of the pit headspace. LESP cannot well predict the mean air velocities and turbulence kinetic

energy in the space next to upwind wall.

(2) The airflow patterns obtained by LESD and LESP were different. Clear vertical air motion

in the top surface of the slot was observed for LESD results. There was not such trend found

for LESP results.

(3) Both the air velocity and NH3 fraction fluctuated weaker for LESP results.

(4) By spectral analysis, LESP was able to capture the entire power spectrum compared with

LESD. A dominant Strouhal number 0.23 was found for LESD results but no dominant

Strouhal number was found for LESP results.

(5) The emission rate and total mass in the pit headspace of NH3 calculated by LESD was

double of those calculated by LESP. Pollutants were confined in the headspace for longer

time by means of using LESP than using LESD.

(6) For both LESD and LESP, turbulence transportation was the dominant removal mechanism

to transport pollutants from the headspace to the free stream.

References

Baik, J.J., Kim, J.J., 1998. A numerical study of flow and pollutant dispersion characteristics in

urban street canyons. Journal of applied meteorology 38, 1576-1589.

Baik, J.J., Kim, J.J., 2002. On the Escape of Pollutants from Urban Street Canyons. Atmospheric

Environment 36, 527-536.

Blocken, B., Stathopoulos, T., Carmeliet, J., 2007.CFD simulation of the atmospheric boundary

layer: wall function problems. Atmospheric environment 41, 238-252.

Bjerg B., Zhang, G., Kai, P., 2008a. Porous media as boundary condition for air inlet, slatted floor

and animal occupied zone in numerical simulation of airflow in a pig unit. AgEng2008

International Conference on Agricultural Engineering, Hersonissos, Crete-Greece.

Bjerg B., Zhang, G., Kai, P., 2008b. CFD investigations of a partly pit ventilation system as method

to reduce ammonia emission from pig production units. The Eighth ASABE International

Livestock Environment Symposium (ILES Ⅷ).

Braam, C. R., Smits, M.C.J., Gunnink, H., Swierstra, D., 1997a. Ammonia Emission from a

Double-sloped Solid Floor in a Cubic House for Dairy Cows. Journal of Agricultural

Engineering Research 68(4), 375-386.

Johnson, G.T., Hunter, L.J., 1998. Urban wind flows: wind tunnel and numerical simulations—a

preliminary comparison. Environmental Modelling and Software 13, 279–286.

Page 143: NEERENGI - Aarhus Universitet

129

Kim, J.J., Baik, J.J., 2003. Effects of inflow turbulence intensity on flow and pollutant dispersion in

an urban street canyon. Journal of Wind Engineering and Industrial Aerodynamics 91, 309-

329.

Launder, B.E., Spalding, D.B., 1974. Numerical computation of turbulent flows. Computers and

Mathematics in Applied Mechanics and Engineering 3, 269-289.

Lee, I.Y., and Park, H.M., 1994. Parameterization of the pollutant transport and dispersion in urban

street canyons. Atmospheric Enviroment 28, 2343-2349.

Liu, C., Leung, D., Barth., M.C., 2005. On the Prediction of Air and Pollutant Exchange Rates in

the Street Canyons of Different Aspect Ratios using Large-eddy Simulation. Atmospheric

Environment 39, 1567-1574.

Morsing, S., Strøm, J.S., Zhang, G., Kai, P., 2008. Scale Model Experiments to Determine the

Effects of Internal Airflow and Floor Design on Gaseous Emissions from Animal Houses.

Biosystems Engineering 99, 99-104.

Neofytou, p., Venetsanos, A.G., Rafailidis, S., Bartzis, J.G., 2006. Numerical investigation of the

pollutant dispersion in an urban street canyon. Environmental Modelling and Software 21,

525–531.

Norton, T., Grant, J., Fallon, R., Sun, D.W., 2010a. Assessing the ventilation effectiveness of

naturally ventilated livestock buildings under wind dominated conditions using computational

fluid dynamics. Biosystems engineering 103 (1), 78-99.

Norton, T., Grant, J., Fallon, R., Sun, D.W., 2010b. A computational fluid dynamics study of air

mixing in a naturally ventilated livestock building with different porous eave opening

conditions. Biosystems engineering 106 (2), 125-137.

Pedersen, S., 2006. Agricultural Best Management Practices in Denmark. In: Proc. Agricultural Air

Quality: State of the Science, 56-67.

Sagrado, A.P.G., van Beeck, J., Rambaud, p., Olivari, D., 2002. Numerical and experimental

modelling of pollutant dispersion in a street Canyon. Journal of Wind Engineering and

Industrial Aerodynamics 90, 321-339.

Saha, C.K., Zhang, G., Ni, J., 2010. Airflow and Concentration Characterization and Ammonia

Mass Transfer Modelling in Wind Tunnel Studies. Biosystems Engineering 107 (4), 328-340.

Solazzo, E., Cai, X., Vardulakis, S., 2009. Improved parameterisation for the numerical modelling

of air pollution within an urban street canyon. Environmental Modelling and Software 24, 381

–388.

Sommer, S.G., Zhang, G., Bannink, A., Chadwick, D., Hutchings, N.J., Misselbrook, T., Menzi, H.,

Ni, J.-Q., Oenema, O., Webb, J., Monteny, G.-J.,2006. Algorithms Determining Ammonia

Emission from Livestock Houses and Manure Stores. Advances in Agronomy 89 (6), 261–335.

Page 144: NEERENGI - Aarhus Universitet

130

Sun, H.W., Keener, H.M., Deng, W., Michel, F., 2004. Development and Validation of 3-D Models

to Simulate Airflow and Ammonia Distribution in a High-rise Hog Building during Summer

and Winter Conditions. Agricultural Engineering International CIGR Journal 6, Manuscript

BC 04 044.

Tsai, M.Y., Chen, K.S., 2004. Measurements and three dimensional modelling of air pollutant

dispersion in an urban street canyon. Atmospheric Environment 38, 5911-5924.

Vardoulakis, S., Fisher, B.E.A., Pericleous, k., Flesca, N.G., 2003. Modelling air quality in street

canyons: a review. Atmospheric Environment 37: 155-182.

Walton, A., Chen, A.Y.S., Yeung, W.C., 2002. Large eddy simulation of pollution dispersion in an

urban street canyon – part II: idealized canyon simulation. Atmospheric Environment 36:

3615-3627.

Wu, W., Zhang, G., Kai, P., 2012a. Emissions of Ammonia and Greenhouse Gases from Two

Naturally Ventilated Dairy Cattle Buildings to Atmosphere – Quantifying the Emissions and

Investigating the Influencing Factors. Manuscript submitted to Atmospheric Environment.

Wu, W., Zhai, J., Zhang, G., Nielsen, P.V., 2012b. Evaluation of methods for determining air

exchange rate in a naturally ventilated dairy cattle building with large openings using

computational fluid dynamics (CFD). Manuscript submitted to Atmospheric Environment.

Wu, W., Zhang, G., Bjerg, B., Nielsen, P.V., 2012d. An assessment of a partial pit ventilation

system to reduce emission under slatted floor-Part 2: Feasibility of CFD prediction using

RANS turbulence models. Computers and Electronics in Agriculture 83, 134-142.

Wu, W., Kai, P., Zhang, 2012d. An assessment of a partial pit ventilation system to reduce emission

under slatted floor-Part 1: Scale Model Study. Computers and Electronics in Agriculture 83,

127-133.

Zhang, G., Strom, J. S., Li, B., Rom, H.B., Morsing, S., Dahl, P., Wang, C., 2005. Emission of

Ammonia and other Contaminant Gases from Naturally Ventilated Dairy Cattle Buildings.

Biosystems Engineering 92 (3), 355-364.

Page 145: NEERENGI - Aarhus Universitet

131

Chapter 8

Numerical Study on the potential of a partial pit ventilation system to reduce ammonia emissions from a naturally ventilated cattle building

Page 146: NEERENGI - Aarhus Universitet

132

Abstract

The hypothesis of this work is to apply a partial pit ventilation system with an air purification unit

to reduce the ammonia emissions from naturally ventilated cattle buildings. Computational fluid

dynamics (CFD) was used to investigate such a potential. The results showed that a pit exhaust with

a capacity of 37.3 m3 h-1 HPU-1 can reduce ammonia emission only by 3.16% compared with the

case without pit ventilation when the external wind was 4.2 m s-1. When the external wind was

decreased to 1.4 m s-1 and the sidewall opening area were reduced to half, such a pit ventilation

capacity can reduce ammonia emission by 85.2%. The utilization of pit ventilation system must be

integrated with the control of the natural ventilation rates of the building.

Key words: CFD; pit ventilation; ammonia emission; livestock building

Page 147: NEERENGI - Aarhus Universitet

133

8.1. Introduction

A partial pit ventilation system with an air purification unit has been known as an efficient approach

to reduce ammonia and odour emissions from pig production units (Bjerg et al., 2008a; Saha et al.,

2010). Bjerg et al. (2008a) investigated a partly pit ventilation system to reduce ammonia emission

from mechanically ventilated pig production units by using computational fluid dynamics (CFD);

they showed that evacuating and cleaning of 10% of the total ventilation capacity from the pit may

reduce the ammonia emission of the system by 73%, and the ammonia concentration in the room

was significantly reduced. Saha et al. (2010) studied the effects of a partial pit ventilation system on

ammonia emission from a fattening pig room with a diffuse ceiling inlet, a ceiling-roof-top air

exhaust and a pit exhaust; the pit exhaust operated with only 10% of the total ventilation capacity

and the ceiling-roof-top exhaust operated as the major ventilation unit regulated according to the

thermal conditions in the room; they concluded that reductions in the ammonia emission of 37-53%

might be achieved by using an additional air cleaning system.

The potential of a partial pit ventilation system to reduce ammonia emissions is seldom tested

in a naturally ventilated cattle building. Bjerg and Andersen (2010) developed a CFD model to

investigate the potentials of a pit ventilation system to abate ammonia emission from a naturally

ventilated cattle building; the results indicated that a pit ventilation system with a capacity of 80 m3

h-1 HPU-1 had the potential to reduce ammonia emission of at least 30%. However, the CFD model

was 2-D and not validated with measurement. Wu et al. (2012a) used a 1:2 scale model of manure

pit section to study the efficiency of a partial pit ventilation to reduce gas emissions; the experiment

was carried out in a wind tunnel. More than 50% of the tracer gas can be removed by the pit

ventilation system. However, the airflow in the wind tunnel cannot represent the wind condition in a

naturally ventilated livestock building.

The objective of this work was to apply CFD methods to investigate the potential of a partial

pit system to reduce ammonia emissions from a full scale and naturally ventilated dairy cattle

building.

8.2. Material and methods

8.2.1. Geometry and mesh

The cubical dairy cow building had a length of 114.0 m, a width of 36.0 m and a height of 11.75 m

(Fig. 8.1a). The detailed description of the naturally ventilated building can be found in the work of

Wu et al. (2012b), in which the building was numbered 2. An office room was adjacent to the

building. Another two neighbour buildings were also included in the simulation. A milking parlour

(Fig. 8.1b) was located in the middle section of the room. The building included four 42.9 m long

slatted floor aisles (Fig. 8.1b). Under the slatted floors were the slurry pits and the pit exhaust

systems were installed to pit headspaces.

Page 148: NEERENGI - Aarhus Universitet

134

Hexahedral mesh was constructed to the entire computational domain (Fig. 8.1c). After a test

of mesh convergence, 1887447 cells were proved to be the optimum grid number for simulations.

a c

b Fig. 8.1 The CFD model and mesh

8.2.2. Boundary conditions

The prescribed boundary conditions are given in Fig. 8.2.

8.2.2.1. Climatic data

The investigation was based on a summer period during Jun 20th and Jul 14th, 2011. The average

wind speed at 10 m height was 4.2 m s-1 and the frequent wind direction was 225° (see Fig. 8.1).

But to study the influence of the external wind speeds, some simulations were carried out at 1.4 to

9.8 m s-1 which covers most levels of the wind speeds during the investigation time. Table 1

Page 149: NEERENGI - Aarhus Universitet

135

provides the reference wind speeds at 10 m height for different simulations cases. The above wind

speeds with a power law function (Norton et al., 2012a) was applied to the velocity inlet (Fig. 8.2)

of the computational domain. The average temperature during the period was 17 °C. The estimation

of thermal conditions for walls in Fig.2 was based on the work of Norton et al. (2012b).

Fig. 8.2 Boundary conditions for the CFD model

8.2.2.2. Pit ventilation capacity

During the investigation time, the house held 128 (Holstein Frisian) dairy cows with an average

weight of 625 kg on average. According to the survey of Seedorf et al. (1998), the mean natural

ventilation rates were 0.77 m3 h-1 kg-1 for dairy cows in cubical houses in Denmark during summer

time. The ventilation rate was equivalent to 373 m3 h-1 HPU-1. 10%, 20% and 30% of the mean

natural ventilation rates were taken as the tested pit ventilation rates, which were about 37.3, 74.5

and 112 m3 h-1 HPU-1. The three associated air velocities 0.06, 0.12 and 0.18 m s-1 for pit exhaust

openings were specified to the velocity outlet in Fig. 8.2 and the airflow direction was normal to the

outlet.

8.2.2.3. Animal Occupied Zone (AOZ) and slatted floor

Heat release from animals was estimated using the method proposed in CIGR (2002). The sensible

heat production was 1291 W cow-1. The resistance introduced by cows was considered as pressure

drop. AOZ was treated as porous media (Bjerg et al., 2008b). The resistance coefficients were 7.71

m-2 for viscous force and 0.06 m-1 for inertial force calculated by Wu et al. (2012c).

Slatted floor was also tackled as porous media. The assumed opening areas were 15%. The

resistance coefficients for 15% opening area were 10000 m-2 for viscous force and 50 m-1 for

Page 150: NEERENGI - Aarhus Universitet

136

inertial force (Bjerg et al., 2008b). Another two slatted floor opening ratios 30% and 7.5% were also

used in this study to investigate the effect of floor openings ratios.

8.2.2.4. Ammonia release

The total ammonia emission rate was generally an unknown parameter. Hence, constant ammonia

mass fraction was specified to the emission surface. Braam et al. (1997) reported about 60% of the

total ammonia emission originates from slatted floor surface, whereas 40% was released from the

slurry pit. The CFD model should consider the both release sources. However, the slatted floor was

treated as porous media and it was impossible to prescribe a constant mass fraction to the slat

surfaces. Therefore, the ammonia was assumed to originate only from slurry pit and the emission

surfaces were marked in Fig. 8.2.

8.2.3. Turbulence models and numerical methods

Wu et al. (2012d) reported that it was feasible to use RSM turbulence models to predict the removal

capability of a partial pit ventilation system. Hence, the widely RSM was adopted in this work.

Commercial software Fluent 12.0 was used to solve the equations on the basis of finite volume

method. All the velocity and turbulence terms for convection were approximated using second order

upwind scheme. Diffusion terms were discretised using central difference scheme. SIMPLE method

was employed for the pressure-velocity correction.

8.2.4. Validation of the CFD model

The measured air velocities by Wu et al. (2012a) were used to validate the CFD model. Air

velocities were measured by 7 ultrasonic anemometers – WindMaster (Gill instruments Ltd,

Hampshire, UK). An ultrasonic anemometer was placed 10 m above the ground to monitor the

external wind speeds. Air velocities inside the buildings were recorded at 6 positions: two were near

one sidewall openings; two were placed in the centre of the buildings and near ridge openings; two

were near another sidewall openings. Meanwhile, velocity profiles were measured in the animal

occupied zones (AOZ) and above feeding aisle. The measured positions and details were not

repeated here and can be found in the work of Wu et al. (2012b).

8.3. Results

8.3.1. Model validation

Fig. 8.3 compares the numerical and experimental air velocity magnitudes at different positions

(Wu et al., 2012a) in the building. The RMSD between simulated and measured mean air velocities

was 0.32. Good agreements were found at the six positions. Fig. 8.4 compares the numerical and

measured air velocity profiles under position D and A (Wu et al., 2012a). Big deviation existed

between the simulated and measured velocity profile. The simulated air velocities were lower than

the minimum values of measurement below 2 m above the ground.

Page 151: NEERENGI - Aarhus Universitet

137

Fig. 8.3 Comparison of numerical (filled bar) and experimental (empty bar) air velocity magnitudes

at different positions (Wu et al., 2012a) in the building

(a) (b) Fig. 8.4 Comparison of numerical (dashed line) and experimental (solid line) air velocity profiles

under the position (Wu et al., 2012a): (a) D and (b) A

8.3.2. Potential of pit ventilation system to reduce ammonia

Table 8.1 shows the reduction of ammonia emission. Case 0 in a situation without pit ventilation,

with fully open door and sidewall openings was taken as the reference case. The ammonia reduction

was based on case 0. Discharge of 37.3 m3 h-1 HPU-1 polluted air through the pit exhaust system can

only lead to ammonia emission reduction by 3.16%. By tripling the pit ventilation rate, only 6.07%

more emission was abated. While the sidewall openings were constrained to 50% of the fully open,

a reduction of ammonia by 43.1% can be achieved compared with Case 0. Pit ventilation with a

capacity of 37.3 m3 h-1 HPU-1 combined with reducing half opening area of the sidewall openings

Page 152: NEERENGI - Aarhus Universitet

138

was able to reduce ammonia emission by 46.0% compared with Case 0. However, compared with

Case 5, pit ventilation abated ammonia emission by 2.9%. The major reduction was indeed

achieved by closing the sidewall openings. When the pit ventilation rate was tripled to 112 m3 h-1

HPU-1, 5.6% more ammonia emission was reduced compared with pit ventilation rate 37.3 m3 h-1

HPU-1. Increasing (Case 10) or decreasing (Case 9) slatted floor opening ratio did not lead to

ammonia reduction in current simulations.

Table 8.1. Reduction of ammonia emission compared with the reference case 0 at the external wind

speed 4.2 m s-1 and wind direction 225°.

Cases pit ventilation rate Door sidewall openings Slatted floor Reduction of ammonia

(m3 h-1 HPU-1) (% of fully open) (% of fully open) opening ratio (%) emission (%)

0 0 100 100 15 0

1 37.3 100 100 15 3.16

2 74.5 100 100 15 6.21

3 112 100 100 15 9.23

4 0 0 100 15 7.57

5 0 100 50 15 43.1

6 37.3 0 50 15 46

7 74.5 0 50 15 48.9

8 112 0 50 15 51.6

9 0 100 100 7.5 0

10 0 100 100 30 0

Fig. 8.5 shows the estimated reduction of ammonia emission as function of external wind

speeds at pit ventilation rate 37.3 m3 h-1 HPU-1 and 50% of the fully open of the sidewall openings.

The pit ventilation system was very efficient at low wind speeds (1.4 m s-1). Compared with Case 0,

ammonia reduction by 85.2% was achieved. Compared with Case 5, pit ventilation system had

ability to reduce 42.2% ammonia emission at low wind speeds.

Fig. 8.5 Estimated reduction of ammonia emission as function of the external wind speeds, at a pit

ventilation rate of 74.5 m3 h-1 HPU-1 and sidewall openings of 50%

Page 153: NEERENGI - Aarhus Universitet

139

8.4. Discussion

8.4.1. Model validation

CFD underestimated the air velocities below 2 m above the floor. The air velocities within this

height were influenced by the AOZ including the existence of partitions, the distribution and

movement of animals. But the influence cannot be fully considered due to the complexity of the

geometrical details. Therefore, AOZ was treated as porous media. The use of porous media may

lead to a difference between simulated and measured air velocities below 2 m. Moreover, the

external wind was unsteady during the measurement. However, the simulation was based on the

mean air velocity and was carried on in a steady situation. The fluctuation of the external wind

might also be responsible for the difference of the numerical and experimental velocity profiles.

The AOZ and the slatted floor were treated as porous media for the purpose of geometry

simplification. Therefore, the effect of slatted floor opening ratio on the ammonia emission was

considered by changing the resistance of porous media. Case 9 and Case 10 did not show a

reduction or an increase in ammonia emission compared with Case 0 by decreasing or increasing

the resistance of porous media. Wu et al. (2012a, d) reported that decreasing the slatted floor

opening ratio can lead to a significant reduction in gas emission. The discrepancy between the

current results and the former research could originate from treating the slatted floor as porous

media. Wu et al. (2012e) reported the difference between simulating slatted floor directly and

treating it as porous media. However, it was difficult to simulate the slatted floor directly in this

simulation. The uncertain of treating slatted floor as porous media cannot be analyzed. In future

work, the uncertain of using porous media should be investigated and the solutions to improve the

accuracy of treating slatted floor as porous media should also be proposed.

8.4.2. Potential of pit ventilation system to reduce ammonia

When the external wind speed was 4.2 m s-1 and the sidewall openings were fully open, the three

designed pit ventilation rates seemed to remove limited ammonia emission (below 10%). The

ventilation rate calculated by CFD simulation in Case 0 was 4485 m3 h-1 HPU-1. The designed pit

ventilation rates were only 0.83%, 1.66% and 2.49% of the estimated natural ventilation rates by

CFD. When the external wind speed was 1.4 m s-1 and the sidewall openings were 50% of the fully

open, the calculated ventilation rate for the building was 663 m3 h-1 HPU-1. The pit ventilation

system with a capacity of 37.3 m3 h-1 HPU-1 (5.6% of the calculated ventilation rate) can reduce

ammonia emission by 85.2% compared with Case 0 and by 42.2% compared with Case 5. This

indicates that utilization of pit ventilation system must be integrated with the control of the natural

ventilation rates of the building. When the building was over-ventilated (Case 0 in this study), the

pit ventilation system might not take good effect. But when the sidewall opening area was reduced

according to external wind and the natural ventilation was kept in a level lower than 663 m3 h-1

Page 154: NEERENGI - Aarhus Universitet

140

HPU-1, ammonia emission from the building can be significantly reduced by the pit ventilation

system.

8.5. Conclusion

Computational fluid dynamics (CFD) was used to investigate the potential of a partial pit ventilation

system with an air purification unit to reduce ammonia emission from naturally ventilated cattle

buildings. Following conclusions can be drawn. A pit exhaust with a capacity of 37.3 m3 h-1 HPU-1

can reduce ammonia emission only by 3.16% compared with the case without pit ventilation when

the external wind was 4.2 m s-1. When the external wind was decreased to 1.4 m s-1 and the sidewall

opening area were reduced to half, such a pit ventilation capacity can reduce ammonia emission by

85.2%. The utilization of pit ventilation system must be integrated with the control of the natural

ventilation rates of the building.

References

Bjerg B., Andersen M., 2010. Numerical simulation of a pit exhausts system for reduction of

ammonia emission from a naturally ventilated cattle building. XVIIth world congress of the

international commission of agricultural and biosystems engineering (CIGR). Quebec City,

Canada.

Bjerg B., Zhang, G., Kai, P., 2008a. CFD investigations of a partly pit ventilation system as method

to reduce ammonia emission from pig production units. The Eighth ASABE International

Livestock Environment Symposium (ILES Ⅷ)

Bjerg B., Zhang, G., Kai, P., 2008b. Porous media as boundary condition for air inlet, slatted floor

and animal occupied zone in numerical simulation of airflow in a pig unit. AgEng2008

International Conference on Agricultural Engineering, Hersonissos, Crete-Greece.

Braam, C. R., Smits, M.C.J., Gunnink, H., Swierstra, D., 1997a. Ammonia Emission from a

Double-sloped Solid Floor in a Cubic House for Dairy Cows. Journal of Agricultural

Engineering Research 68(4), 375-386.

CIGR, 2002. Report of Working Group on Climatization of Animal Houses – Heat and Moisture

Production at Animal and House Levels. CIGR Section II, Commission International Du

Genie Rural (International Commission of Agricultural Engineering).

Launder, B.E., Spalding, D.B., 1974. The numerical computational of turbulent flows. Computer

Methods in Applied Mechanics and Engineering 3, 269-289.

Norton, T., Grant, J., Fallon, R., Sun, D.W., 2010a. Assessing the ventilation effectiveness of

naturally ventilated livestock buildings under wind dominated conditions using computational

fluid dynamics. Biosystems engineering 103 (1), 78-99.

Page 155: NEERENGI - Aarhus Universitet

141

Norton, T., Grant, J., Fallon, R., Sun, D.W., 2010b. A computational fluid dynamics study of air

mixing in a naturally ventilated livestock building with different porous eave opening

conditions. Biosystems engineering 106 (2), 125-137.

Saha, C.K., Zhang, G., Kai, P., Bjerg B., 2010. Effects of a partial pit ventilation system on indoor

air quality an ammonia emission from a fattening pig room. Biosystems Engineering 105 (3),

279-287.

Seedorf, J., Hartung, J., Schroder, M., Linkert, K.H., Pedersen, S., Takai, H., Johnsen, J.O., Metz,

J.H.M., Groot Koerkamp, P.W.G., Uenk, G.H., Phillips, V.R., Holden, M.R., Sneath, R.W.,

Short, J.L.L., White, R.P., Wathes, C.M., 1998. A Survey of Ventilation Rates in Livestock

Buildings in Northern Europe. Journal of Agricultural Engineering Research 70, 39-47.

Wu, W., Kai, P., Zhang, 2012a. An assessment of a partial pit ventilation system to reduce emission

under slatted floor-Part 1: Scale Model Study. Computers and Electronics in Agriculture 83,

127-133.

Wu, W., Zhang, G., Kai, P., 2012b. Ammonia and Methane Emissions from Two Naturally

Ventilated Dairy Cattle Buildings and the Influence of climatic Factors on Ammonia

Emissions. Accepted by Atmospheric Environment.

Wu, W., Zhai, J., Zhang, G., Nielsen, P.V., 2012c. Evaluation of methods for determining air

exchange rate in a naturally ventilated dairy cattle building with large openings using

computational fluid dynamics (CFD). Submitted to a peer review journal.

Wu, W., Zhang, G., Bjerg, B., Nielsen, P.V., 2012d. An assessment of a partial pit ventilation

system to reduce emission under slatted floor-Part 2: Feasibility of CFD prediction using

RANS turbulence models. Computers and Electronics in Agriculture 83, 134-142.

Wu, W., Zong, C., Zhang, G., 2012e. Large eddy simulation of airflow and pollutant transport

under slatted floor of a slurry pit model. Submitted to a peer review journal.

Page 156: NEERENGI - Aarhus Universitet

142

Chapter 9

General discussion and conclusions

Page 157: NEERENGI - Aarhus Universitet

143

9.1. Introduction

According to the objectives, this thesis is subsequently divided into two sections: predicting

emissions from naturally ventilated dairy cattle buildings and establishing a systematic approach to

reduce the emissions.

The first section includes Chapter 2, 3, and 4. Chapter 2 is aimed at predicting gas emissions

from naturally ventilated cattle buildings. Therefore, gas concentrations and their influencing

parameters like air velocity and temperature were measured in two farm buildings. The emission

rates of ammonia and methane as well as ventilation rates can thus be calculated by CO2 production

model (CIGR, 2002; Zhang et al., 2005; Feidler and Müller, 2011). The influence of air velocity

and temperature on ammonia emissions was analyzed by multiple linear regressions. Both the

ventilation rate and the gas concentration distribution were not only affected by air velocities but

also by turbulence. The turbulence characteristics were discussed in an individual chapter –

Chapter 3. The air velocity time series for turbulence calculation were measured by ultrasonic and

the setup of the measurement in the two farm buildings was presented in Chapter 2. A discussion of

the uncertainty of CO2 production model was barely found in literature. Hence, it was investigated

in Chapter 4 using computational fluid dynamics (CFD) by comparisons with other methods.

Moreover, one remained challenge of using CO2 production model was the determination of the

sampling positions for the outlet gas concentration, which was termed as representative gas

concentration in this thesis. The representative positions to sample outlet gas were thus proposed.

In the second section including Chapter 5, 6, 7 and 8, efforts were put on developing a partial

pit ventilation system to mitigate the emissions. Scale model of slurry pit was constructed and wind

tunnel tests were conducted to study the contaminant removal efficiency of such a system under

varied air speeds above floor, and ventilation configurations. The laboratory experiment and

research results were presented in Chapter 5. The laboratory tests need to be verified by full scale

investigations. Partial pit ventilation systems have not been applied to naturally ventilated dairy

cattle buildings. Therefore, the full scale investigations had to be carried out by CFD simulations.

Prior to full scale simulations, Chapter 6 stated the possibility of using RANS turbulence models to

perform investigations on pollutant transportation from slurry pit configured with mechanical pit

exhaust systems. The simulations were calibrated by the laboratory measurements described in

Chapter 5. The geometrical details of slatted floor generally cannot be modelled directly in a full

scale simulation (Bjerg et al., 2008). Bjerg et al. (2008) dealt with the slatted floor as porous media.

However, the uncertainty of treating slatted floor as porous media and modeling them directly was

not analyzed in literature. The uncertainty was vital even if the porous media had to be used. The

difference of pollutant transportation by treating slatted floor as porous media and modeling them

directly was studied by CFD in Chapter 7. Simulations were validated by air velocities measured

by Laser Doppler anemometer in a wind tunnel. After finishing the preparation work in Chapter 6

Page 158: NEERENGI - Aarhus Universitet

144

and 7, the full scale simulation was eventually performed in Chapter 8. Pit ventilation systems

were tested in different conditions.

9.2. Determination of ventilation and gas emission rates

To predict ammonia and methane emissions, gas concentrations were measured inside and outside

two naturally ventilated buildings (Chapter 2). The emission rates of ammonia and methane as well

as ventilation rates can thus be calculated by CO2 production model (CIGR, 2002; Zhang et al.,

2005; Feidler and Müller, 2011). As the CO2 was assumed to come from animals, and the

production was known based on the cow weight and milk yield. The inputs for CO2 production

model to calculate ventilation rates were background and outlet CO2 concentrations. The

background concentration was sampled outside the buildings. The outlet concentration was sampled

near the side wall and ridge openings inside the buildings. The gas concentration near the openings

was assumed to be able to represent the outlet concentration, which was termed as representative

gas concentration in this thesis. The setup of the measurements was supported by the research of

Demmers et al. (1998), which showed that the mean tracer gas concentration measured in the

openings was better to represent the outlet gas concentration than that measured in the building. The

predicted ventilation rates as well as background and outlet concentrations of associated gas were

used to calculate emission rates were the. Therefore, the accuracy of the quantification of the gas

emission rates using CO2 production model was highly depended on the performance of the CO2

production model and the sampling positions of outlet gas concentrations.

In order to evaluate the performance of the CO2 production model to predict ventilation rates

or emission rates, the method should be validated against or compared to other methods. Natural

ventilation rates can be decided using three existing methods – integrating volume flow rates (VFR)

across all the openings, tracer gas decay (TGD) and constant tracer gas (CTG). CO2 production

model can be grouped into CTG. Demmers et al. (1998) showed that VFR was impossible to

implement in reality due to the large openings and small pressure difference across each opening.

TGD method cannot discover the ventilation variability during the decay process. Moreover, the

external wind conditions could vary rapidly from one hour to next hour. Hence, it is hard to

associate the ventilation rates estimated by TGD to the values estimated by CTG subjecting to the

same external wind conditions. Due to the above limitations, computational fluid dynamics (CFD)

has become a surge to assess the performance of CO2 production model against VFR and CTG. All

the three methods can be easily implemented in CFD. To find the sampling positions of outlet gas

concentrations via experiment, more instruments and measurement systems are needed. It will also

be very expensive, time consuming and laborious. But the same things can be easily performed in

CFD simulations. In addition, CFD modelling can exploit insights into the airflow patterns,

temperature and concentration distribution governing the gas emissions. Hence, CFD is used as a

tool to assess the uncertainty in quantifying gas emission rates using CO2 production model.

Page 159: NEERENGI - Aarhus Universitet

145

The CFD simulation results were validated by air velocity and CO2 concentration data

(Chapter 4). While the VFR was set as reference method, TGD demonstrated its good agreements

with VFR. CO2 production model using the mean concentration of the position near downwind

sidewall opening also agreed with VFR over a wide range of external wind speeds and directions.

CO2 production model using the mean concentration of the entire room failed to predict consistent

ventilation rates with VFR and TGD. Therefore, the sampling positions are vital to the accuracy of

the CO2 production model.

In order to find the optimum gas sampling positions four gas concentrations were compared

which were (1) the mean gas concentration of the entire room calculated by CFD; (2) the mean gas

concentration of three positions respectively near upwind sidewall opening, ridge opening and

downwind sidewall opening; (3) the mean gas concentration of two positions respectively near

ridge opening and downwind sidewall opening; (4) the mean gas concentration of the position near

downwind sidewall opening. The mean gas concentration of the entire room failed to represent that

in the exit air. The result agreed with the work of Demmers et al. (2000), which also reported large

error in ventilation rate estimate using internal sampling points; they also indicated that no obvious

zones or locations, which offered a representative concentration, could be identified within the

building section. The mean gas concentration of the position near downwind sidewall opening was

proved to best represent the outlet gas concentration. If CO2 production model was used to calculate

ventilate rates or emission rates from naturally ventilated livestock buildings, the gas sampling

positions should be adjacent to or even located in the openings. The maximum gas concentration at

different positions will be the representative gas concentration.

9.3. Climatic factors on gas emissions

Emission issues including odour, gas and particle from a naturally ventilated livestock building

were controlled by the climatic factors such as wind speed, direction, temperature and turbulence as

well as building components such as ventilation opening areas, positions and slatted floor. This

information is also essential to model naturally ventilated cattle buildings using CFD.

9.3.1. Air velocity and temperature

The climatic data was collected synchronously with the concentration data as presented in Chapter

2. The correlation between climatic factors and ammonia emissions was analyzed by multiple linear

regressions. Wind speeds and air temperature showed significant and positive correlation with

ammonia emission. The similar conclusions were confirmed by literature (Snell et al., 2003; Zhang

et al., 2008; Schrade et al., 2012). The wind speeds and directions can be monitored by ultrasonic

placed outside and inside the buildings. The external wind speeds and directions can then be

transformed to velocity profiles prescribed as the boundary conditions of the CFD model. Since the

temperature was identified as an import factor influencing emissions, the temperature of external air

Page 160: NEERENGI - Aarhus Universitet

146

and wall surfaces of building should be defined. No measured wall temperature was available. The

associated values were derived by heat balance equation, which was described in Chapter 4.

9.3.2. Air turbulence characteristics

The turbulence characteristics in two naturally ventilated cattle buildings were analyzed in detail in

terms of autocorrelation, kinetic energy, turbulence energy dissipation rate, integral time and length

scale as well as Kolmogorov microscale of length. Spectral analysis was also applied to the time

series of external and internal air velocities in order to investigate the coherency of turbulence

structures. An outstanding outcome of the air turbulence analysis was the irregular distribution of

all the statistical descriptors except that the kinetic energy. When the external wind speed was

increased, the internal kinetic energy at all the measured internal positions was increased

accordingly. The purpose of the analysis of air turbulence in the buildings was to find the

correlation between gas emission process and turbulence. Although no simple trend or correlation

was found among the different statistical descriptors, the calculation of turbulence parameters can

be useful in the following manners:

(1) CFD is a major tool in this research. The calculation of turbulence length scale, kinetic

energy and dissipation rate can provide appropriate boundary conditions for the CFD

model.

(2) Spectral analysis might be used to build mathematical models to predict natural

ventilation rates (Lay and Bragg, 1998); Velocity time series might be transformed to

frequency domain; a mathematical model to ventilation rate based on velocity data could

be established in frequency domain.

9.4. Potential of partial pit ventilation systems to reduce emission

9.4.1. Laboratory measurements

In a mechanically ventilated livestock building, an air purification unit can be mounted to the

exhaust duct and reduce the gas emissions. Unlike the mechanically ventilated buildings, it is

difficult to collect the exhausted air for further air cleaning in a naturally ventilated livestock

building. However, a concept of partial pit ventilation may be adapted into the housing system. A

mechanical ventilation system can be connected to the headspace of the slurry pit and removed a

part of the most polluted air under the slatted floor. The exhausted air was then purified by filtering

apparatus. The hypothesis was firstly validated in a laboratory condition using a scale model. The

detailed setup was described in Chapter 5. Partial pit ventilation was approved to be able to remove

a large portion of polluted gases under the slatted floor in the scale model measurement. More than

50% of the contaminants can be removed by the pit ventilation system. The laboratory test may

exaggerate the function of the pit ventilation system compared with the practical situation. Braam et

al. (1997) reported about 60% of the total ammonia emission originated from slatted floor surface,

Page 161: NEERENGI - Aarhus Universitet

147

whereas 40% was released from the slurry pit in a naturally ventilated cow building. However, the

scale model measurement in this study did not consider the emission from the surface of the slatted

floor. Another deficiency of the scale model measurement was that the airflow distribution above

the slatted floor in the wind tunnel may not represent that in a practical building.

9.4.2. Using CFD to investigate the performance of pit ventilation system

Different RANS turbulence models were tested and an underestimation of removal ratio was

noticed for all the turbulence models. The transition SST k turbulence model underestimated

the removal ratio more than 8.1%. The three k models predicted similar removal ratios. The

deviations from the measured values were larger than 5%. The best agreement between CFD

simulation and measurement was achieved by RSM turbulence model. The deviation was less than

2%. Successful modelling of the flow over the pit headspace depended on the ability of the

associated turbulence models to simulate the separation of fluid. The eddy-viscosity turbulence

models like k models can sometimes over-predict the turbulence kinetic energy and were not

sensitive to the interaction between streamline curvature and turbulence anisotropy. RSM, on the

other hand, accounted for several turbulence features like streamline curvature. This may explain

why the RSM showed a better performance.

Removal ratios based on CFD simulations were very close to measured values in most cases.

It was feasible to use RANS turbulence models to assess the ability of a partial pit ventilation

system to reduce emission under slatted floor. However, the study also showed that the gas

transportation from the slurry pit to the free stream was dominated by the turbulence diffusion.

Johnson and Hunter (1998) suggested that transient flow was an important factor to model flow and

pollutant transportation dominated by turbulence diffusion.

9.4.3. Modeling of slatted floor

Slatted floors were the major component of the housing system with slatted floors. Except that

slatted floors were the emission source, they were also the interfaces of the polluted air in the pit

headspace and the air in the room. The boundary layer around the slatted floors governed the

pollutant transportation from the pit headspace to the room space. Therefore, the accuracy of

modeling of slatted floor decided the accuracy of simulating gas emissions from naturally ventilated

cattle buildings. In practical buildings, hundreds of slats were used and a detailed simulation of the

geometry using CFD was not realistic. They were generally treated as porous media. However, the

uncertainty analysis of using porous media cannot be found in current literature. Although it was

hard to avoid using porous media to simulate full scale buildings with slatted floor, an investigation

of modeling slatted floor directly and treating them as porous media was particularly useful. To

validate the CFD model, a 1:8 scale pit model was constructed in a wind tunnel. To consider the

transient effect of turbulence diffusion, large eddy simulation was adopted to study the ammonia

Page 162: NEERENGI - Aarhus Universitet

148

transportation. To tackle the involvement of the slatted floor, two CFD models were established:

modelling slatted floors directly with geometrical details (LESD) and treating them as porous media

(LESP). LESP and LESD estimated similar mean air velocities and turbulence kinetic energy in the

core of the pit headspace. Large discrepancy existed next to the upwind wall between both methods.

Some other major differences between the LESP and LESD were:

(1) A dominant Strouhal number 0.23 was found for LESD results but no dominant strouhal

number was found for LESP results.

(2) The emission rate and total mass of NH3 in the pit headspace calculated by LESD was

double of those calculated by LESP. Pollutants were confined in the headspace for longer

time by means of using LESP than using LESD.

(3) The airflow patterns were different between LESP and LESD. Clear vertical air motion in

the top surface of the slot was observed for LESD results but not for LESP.

The current simulation only considered the case when the slats were oriented parallel to the

flow direction. However, the difference of LESP and LESD may be related to the orientation of the

slats. For instance, if the slats were perpendicular to the flow direction, the difference might be

more significant. From the point view of emission rates and the flow patterns, porous media may be

not sufficient to represent the slatted floor. New methods may be needed to simplify and simulate

the slatted floors.

9.4.4. Full scale simulation

CFD has been proved to be a feasible approach to study the ventilated flow in livestock buildings

(Norton et al., 2010). The full scale investigation of the potential of partial pit ventilation systems to

reduce emissions was completed by CFD. Although treating slatted floor as porous media might

introduce uncertainties to the results, using porous media cannot be avoided due to the complicity of

the building geometry and the computational capacity. If the livestock building was over-ventilated,

pit ventilation with 10% capacity of the designed room ventilation rates may not be able reduce

emissions efficiently. In this study, pit ventilation systems can be efficient only if the curtains of

sidewall openings were regulated according to the external wind conditions. In practice, the

sidewall openings are usually fully open during summer in naturally ventilated cattle buildings in

order to remove the heat and the humidity. Under this circumstance, the buildings in windy zones

(Denmark) are generally over-ventilated. During winter, the sidewall openings might be closed to

provide warmer indoor climate for cows and ventilation might be dominated by stack effect. In this

situation, pit ventilation systems can be very efficient to reduce the gas emissions.

9.5. Perspectives

The first part of the thesis solved the uncertainties concerning quantification of ventilation and

emission rates using CO2 production model. As a result, a major outcome was to find the optimum

Page 163: NEERENGI - Aarhus Universitet

149

gas sampling positions to represent the outlet concentration. The optimum gas sampling positions

were determined by CFD. In future, different sampling systems and positions should be setup in the

field measurement. Therefore, the optimum positions determined by measurement can be compared

with those found by CFD. Further study may also be focused on proposing new methodology to

determine natural ventilation rate instead of using CO2 production model. The new methodology to

determine natural ventilation rates may be complemented by Fourier transform of data in time

domain into frequency domain. A mathematical model on ventilation rates might be developed in

frequency domain by necessary parameters. In addition, the emission inventories in this study were

campaigned in two naturally ventilated dairy cattle buildings. More measurement of gas

concentration data were needed to add to the database for national emission inventories.

The second part of the thesis was devoted to applying pit ventilation systems to reduce

emissions. It was first tested in a scale model, and then the full scale investigation was done by CFD

simulations. There existed many uncertainties introduced by simplification building components

and animals. Animals were treated as porous media in this study according to Bjerg et al. (2008). To

compare the effect of animals, different simplification methods like treating them as blocks can be

the potential work. In chapter 7, slatted floor was modeled directly and as porous media.

Simulations on different orientations of slats relative to the flow direction were necessary to study

the uncertainty of using porous media. Moreover, new numerical methods to tackle slatted floor

may be proposed to model this essential zone. The current simulations were conducted in steady

conditions. As it was pointed out in literature, transient flow was an important feature of pollutant

transportation. Unsteady simulations of natural ventilation across livestock buildings can be very

helpful to study the potential of partial pit ventilation systems to reduce emissions.

9.6. General conclusions

The following general conclusions were drawn from this thesis:

The NH3 emission rates varied from 32-77 g HPU-1 d-1 in building 1 and varied from 18-30 g

HPU-1 d-1 in building 2. The average emission of CH4 was 290 and 230 g HPU-1 d-1 from

building 1 and 2, respectively. Diurnal pattern was found for NH3 and CH4 emission rates.

From multiple linear regression models, there was a significant linear relationship between

NH3 emission rates and climatic factors including the external wind speed as well as the air

temperature (P<0.001), but not with the external wind directions (P>0.05).

The distribution of turbulence time and length integral scale was very irregular. The internal

kinetic energy had a positive correlation with the external wind speed. Larger turbulence

energy dissipation rate (0.49 m2 s-3) was found near side openings, where the integral length

scale was smaller. Kolmogorov microscales were quite isotropy for the natural airflow in this

study. Air velocities near downwind side openings possessed higher power spectra at both

Page 164: NEERENGI - Aarhus Universitet

150

low and high frequency, compared with those in the middle section or near upwind side

openings.

Air exchange rates (AER) predicted by integrating volume flow rate (VFR) and tracer gas

decay (TGD) were in good agreement with each other within a large range of wind speeds.

Large difference in AER estimation using constant tracer gas method (CTG) and VFR

indicates that the mean CO2 concentration of the entire room may not represent the outlet

concentration.

The gas sampling positions should be located adjacent to the openings or even in the

openings. To reduce the uncertain introduced by wind direction, all the openings especially of

different azimuths should possess sampling tubes. The maximum gas concentrations in the

different openings could be the optimum value to represent the concentration in the exit air.

Scale model experiment showed that partial pit ventilation was able to remove a large portion

of polluted gases under the slatted floor.

RANS turbulence models especially RSM can be used to predict the removal capability of a

partial pit ventilation system to reduce emission under slatted floor.

Two CFD approaches were used to tackle the involvement of the slatted floor: modelling

slatted floors directly with geometrical details (LESD) and treating them as porous media

(LESP). LESP and LESD estimated similar mean air velocities and turbulence kinetic energy

in the core of the pit headspace. Large discrepancy existed next to the upwind wall between

both methods. The emission rate and total mass of NH3 in the pit headspace calculated by

LESD was double of those calculated by LESP. Pollutants were confined in the headspace for

longer time by means of using LESP than using LESD. The airflow patterns were different

between LESP and LESD. Clear vertical air motion in the top surface of the slot was observed

for LESD results but not for LESP.

A pit exhaust with a capacity of 37.3 m3 h-1 HPU-1 can reduce ammonia emission only by

3.16% compared with the case without pit ventilation when the external wind was 4.2 m s-1.

When the external wind was decreased to 1.4 m s-1 and the sidewall opening area were

reduced to half, such a pit ventilation capacity can reduce ammonia emission by 85.2%. The

utilization of pit ventilation system must be integrated with the control of the natural

ventilation rates of the building.

References

Bjerg B., Zhang, G., Kai, P., 2008b. Porous media as boundary condition for air inlet, slatted floor

and animal occupied zone in numerical simulation of airflow in a pig unit. AgEng2008

International Conference on Agricultural Engineering, Hersonissos, Crete-Greece.

Page 165: NEERENGI - Aarhus Universitet

151

Braam, C. R., Smits, M.C.J., Gunnink, H., Swierstra, D., 1997a. Ammonia Emission from a

Double-sloped Solid Floor in a Cubic House for Dairy Cows. Journal of Agricultural

Engineering Research 68(4), 375-386.

CIGR, 2002. Report of Working Group on Climatization of Animal Houses – Heat and Moisture

Production at Animal and House Levels. CIGR Section II, Commission International Du

Genie Rural (International Commission of Agricultural Engineering).

Demmers, T.G.M., Burgess, L.R., Short, J.L., Phillips, V.R., Clark, J.A., Wathes., C.M., 1998. First

experiences with methods to measure ammonia emissions from naturally ventilated cattle

buildings in the U.K. Atmospheric Environment 32,285-293.

Demmers, T.G.M., Burgess, L.R., Phillips, V.R., Clark, J.A., Wathes., C.M., 2000. Assessment of

Techniques for Measuring the Ventilation Rate, using an Experimental Building Section.

Journal of Agricultural Engineering Research 76,71-81.

Feidler, A.M., Müller, H.J., 2011. Emissions of ammonia and methane from a livestock building

natural cross ventilation. Meteorologische Zeitschrift 20 (1), 059-065.

Johnson, G.T., Hunter, L.J., 1998. Urban wind flows: wind tunnel and numerical simulations—a

preliminary comparison. Environmental Modelling and Software 13, 279–286.

Lay, R. M., Bragg, G. M., 1998. Distribution of Ventilation air – Measurement and Spectral

Analysis by Microcomputer. Building and Environment, 23 (3), 203-213.

Norton, T., Grant, J., Fallon, R., Sun, D.W., 2010a. Assessing the ventilation effectiveness of

naturally ventilated livestock buildings under wind dominated conditions using

computational fluid dynamics. Biosystems engineering 103 (1), 78-99.

Schrade, S., Zeyer, K., Gygax, L., Emmenegger, L., Hartung, E., Keck, M., 2012. Ammonia

Emissions and Emission Factors of Naturally Ventilated Dairy Housing with Solid Floors

and an Outdoor Exercise Area in Switzerland. Atmospheric Environment 47, 183-194.

Snell, H.G.J., Seipelt, F., Van Den Weghe, H.F.A., 2003. Ventilation Rates and Gaseous Emissions

from Naturally Ventilated Dairy Houses. Biosystems Engineering 86 (1), 67-73.

Zhang G; Bjerg B; Strøm JS; Morsing S; Kai P; Tong G; Ravn P, 2008. Emission Effects of Three

different Ventilation Control Strategies - A Scale Model Study. Biosystems Engineering

100(1), 96-104.

Page 166: NEERENGI - Aarhus Universitet

Department of Engineering Aarhus University Edison, Finlandsgade 22 8200 Aarhus N Denmark

Tel.: +45 4189 3000

Wentao Wu, Modelling and reducing gas emissions from naturally ventilated livestock buildings, 2012